WELFARE, THE EARNED INCOME TAX CREDIT, AND …scholz/Teaching_742/Meyer...WELFARE, THE EARNED INCOME...

52
WELFARE, THE EARNED INCOME TAX CREDIT, AND THE LABOR SUPPLY OF SINGLE MOTHERS* BRUCE D. MEYER AND DAN T. ROSENBAUM During 1984 –1996, welfare and tax policy were changed to encourage work by single mothers. The Earned Income Tax Credit was expanded, welfare bene ts were cut, welfare time limits were added, and welfare cases were terminated. Medicaid for the working poor was expanded, as were training programs and child care. During this same time period there were unprecedented increases in the employment and hours of single mothers. We show that a large share of the increase in work by single mothers can be attributed to the EITC and other tax changes, with smaller shares for welfare bene t cuts, welfare waivers, training programs and child care programs. I. INTRODUCTION Between 1984 and 1996, changes in tax and transfer pro- grams sharply increased the incentive for single mothers to work. During this same period, single mothers began to work more as their weekly employment increased by about six percentage points and their annual employment increased by nearly nine percentage points. Other groups, such as single women without children, married mothers, and black men, did not experience similar gains in employment over this period (see Meyer and Rosenbaum [2000a]). These facts lead us to examine whether the changes in tax and transfer programs were responsible for single mothers working more and what changes were the most important. * We thank Joseph Altonji, Timothy Bartik, Rebecca Blank, Janet Currie, Steven Davis, Jeffrey Liebman, Thomas MaCurdy, Leslie Moscow, Derek Neal, Julie Rosenbaum, Christopher Ruhm, John Karl Scholz, seminar participants at the Association for Public Policy Analysis and Management Annual Meetings, University of California at San Diego, the Federal Reserve Board, the Institute for Research on Poverty, the Manpower Demonstration Research Corporation, the University of Maryland, the National Association for Welfare Research and Sta- tistics Annual Meetings, the National Bureau of Economic Research, the National Tax Association annual meetings, Northwestern University, Syracuse University, Texas A & M University, the Urban Institute, and the University of Wisconsin and many others for comments. We have been greatly aided by help with data and program details by Daniel Feenberg, Linda Giannarelli, Laura Guy, Phillip Le- vine, Robert McIntire, Steve Savner, Jill Schield, and Aaron Yelowitz. Brian Jenn and Christopher Jepsen provided excellent research assistance. This research has been partly supported by the Northwestern University/University of Chicago Joint Poverty Center, the Household International, Inc. Chair in Economics and the National Science Foundation (Meyer), and a Research Training Fellowship in Urban Poverty funded by the National Science Foundation and Northwestern University (Rosenbaum). © 2001 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology. The Quarterly Journal of Economics, August 2001 1063

Transcript of WELFARE, THE EARNED INCOME TAX CREDIT, AND …scholz/Teaching_742/Meyer...WELFARE, THE EARNED INCOME...

WELFARE THE EARNED INCOME TAX CREDIT ANDTHE LABOR SUPPLY OF SINGLE MOTHERS

BRUCE D MEYER AND DAN T ROSENBAUM

During 1984ndash1996 welfare and tax policy were changed to encourage work bysingle mothers The Earned Income Tax Credit was expanded welfare benetswere cut welfare time limits were added and welfare cases were terminatedMedicaid for the working poor was expanded as were training programs and childcare During this same time period there were unprecedented increases in theemployment and hours of single mothers We show that a large share of theincrease in work by single mothers can be attributed to the EITC and other taxchanges with smaller shares for welfare benet cuts welfare waivers trainingprograms and child care programs

I INTRODUCTION

Between 1984 and 1996 changes in tax and transfer pro-grams sharply increased the incentive for single mothers to workDuring this same period single mothers began to work more astheir weekly employment increased by about six percentagepoints and their annual employment increased by nearly ninepercentage points Other groups such as single women withoutchildren married mothers and black men did not experiencesimilar gains in employment over this period (see Meyer andRosenbaum [2000a]) These facts lead us to examine whether thechanges in tax and transfer programs were responsible for singlemothers working more and what changes were the mostimportant

We thank Joseph Altonji Timothy Bartik Rebecca Blank Janet CurrieSteven Davis Jeffrey Liebman Thomas MaCurdy Leslie Moscow Derek NealJulie Rosenbaum Christopher Ruhm John Karl Scholz seminar participants atthe Association for Public Policy Analysis and Management Annual MeetingsUniversity of California at San Diego the Federal Reserve Board the Institute forResearch on Poverty the Manpower Demonstration Research Corporation theUniversity of Maryland the National Association for Welfare Research and Sta-tistics Annual Meetings the National Bureau of Economic Research the NationalTax Association annual meetings Northwestern University Syracuse UniversityTexas A amp M University the Urban Institute and the University of Wisconsinand many others for comments We have been greatly aided by help with data andprogram details by Daniel Feenberg Linda Giannarelli Laura Guy Phillip Le-vine Robert McIntire Steve Savner Jill Schield and Aaron Yelowitz Brian Jennand Christopher Jepsen provided excellent research assistance This research hasbeen partly supported by the Northwestern UniversityUniversity of ChicagoJoint Poverty Center the Household International Inc Chair in Economics andthe National Science Foundation (Meyer) and a Research Training Fellowship inUrban Poverty funded by the National Science Foundation and NorthwesternUniversity (Rosenbaum)

copy 2001 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics August 2001

1063

The largest change in the work incentives of single mothersbetween 1984 and 1996 was a tenfold increase in credits throughthe Earned Income Tax Credit (EITC) Only working families(primarily those with children) receive the EITC so its expansionincreased the incentive for single mothers to work We analyzenot only the federal EITC but other federal income tax changesstate income taxes and state EITCs as well The Medicaid pro-gram also greatly expanded during this period Between 1984 and1994 the number of children receiving health coverage throughMedicaid increased 77 percent while the number of coveredadults with dependent children increased 35 percent The expan-sions increased coverage for nonwelfare families with incomesnear the poverty line thus making work more attractive forlow-income single mothers Cash assistance to single parentsthrough Aid to Families with Dependent Children (AFDC) alsochanged quite dramatically over this period Nearly every stateexperimented with changes often under federal waivers of theexisting AFDC rules These changes typically imposed work re-quirements time limits or other measures to encourage singlemothers to work We also investigate the effects of other changesto the AFDC and Food Stamp programs including changes inbenet levels earnings disregards and benet reduction ratesFinally we examine the effects of changes in child care andtraining programs during this period

Our main research strategy identies the effects of thesepolicies on single mothersrsquo labor supply through the differentialtreatment of single mothers and single women without childrenunder welfare and tax laws However the richness of these policychanges allows us to consider additional specications that focuson narrower sources of variation including differences amongsingle mothers in their numbers and ages of children and differ-ences across states in their taxes benets and living costs Thesesources of variation are likely to be unrelated to underlyingdifferences across individuals in their desire to work and thusare likely to be exogenous to labor supply decisions We alsodevelop a new methodology for summarizing the key features ofthe complex nonlinear budget sets created by policies such as theEITC Medicaid and AFDC

Understanding the relationship between the changes in gov-ernment policies and the increases in the labor supply of singlemothers during this period is important for several reasons Firstthese changes in policies provide a plausible source of exogenous

1064 QUARTERLY JOURNAL OF ECONOMICS

variation with which to identify the effects of tax and welfareparameters on labor supply The magnitudes of these effects arekey determinants of the gains or losses from changes in incomeredistribution and social insurance policies

Second understanding the effects of government policiesduring the 1984 ndash1996 period has taken on more importance dueto the passage of the Personal Responsibility and Work Opportu-nity Reconciliation Act of 1996 (PRWORA) In 1997 PRWORAreplaced the main cash assistance program for single mothersAFDC with Temporary Assistance for Needy Families (TANF)The increased state discretion under the new law combined withpolitical changes has led to welfare reform which discourageswelfare receipt and often diverts potential welfare recipients fromtraditional programs These reforms are difcult if not impossibleto characterize using a few variables It is likely that many of thepolicies examined in this paper will be harder and more problem-atic to analyze using post-PRWORA data1

Third there is surprisingly little previous work that esti-mates the effects of the EITC Medicaid or welfare changes onwhether single mothers work The only paper that directly exam-ines how the EITC affects single mothersrsquo labor supply is Eissaand Liebman [1996] which examines the effect of the Tax ReformAct of 19862 In his discussion of the labor supply effects ofMedicaid Moftt [1992] argues that there has been too little workto draw reliable conclusions3 Moftt describes the labor supplyeffect of AFDC as being subject to considerable uncertainty andnotes that the broader labor supply literature has examinedsingle mothers ldquoonly rarelyrdquo4 Dickert Houser and Scholz [1995]argue that this literature provides little guidance as to how theEITC will affect labor market participation and that this omis-sion is especially important because past work suggests that most

1 See Ellwood [2000] National Research Council [1999] and Jencks andSwingle [2000] for related arguments

2 Several papers use labor supply parameters estimated from the negativeincome tax experiments and other sources to simulate the effects of the EITCincluding Hoffman and Seidman [1990] Holtzblatt McCubbin and Gillette[1994] Browning [1995] and Dickert Houser and Scholz [1995] Dickert Houserand Scholz estimate the effect of the after-tax wage and welfare programs onparticipation using a cross section of data from the 1990 panel of the Survey ofIncome and Program Participation (SIPP) They then apply these results tosimulate the effects of the EITC on participation Eissa and Hoynes [1998]examine the effects of the EITC on the labor supply of married couples

3 See Blank [1989] Winkler [1991] and Moftt and Wolfe [1992] in particu-lar The more recent work of Yelowitz [1995] examines the 1988 to 1991 period

4 See Danziger Haveman and Plotnick [1981] and Moftt [1992]

1065THE LABOR SUPPLY OF SINGLE MOTHERS

of the labor supply response is in the work decision rather thanthe hours decision Furthermore there is no work that we areaware of that assesses the overall effect of recent changes intraining and child care programs5 The work on the effects ofwelfare waivers has examined program caseloads rather thanemployment and has reached conicting results6

We examine the major policies affecting the labor supply ofsingle mothers during the 1984 to 1996 period using two datasets the Current Population Survey (CPS) Outgoing RotationGroup Files and the March CPS Files By investigating severalprograms at once using thirteen years of individual data weaccount for their separate effects and we can directly comparethe programs using the same sample time period and methodsOur approach improves on the common past research strategy ofexamining changes in one of these policies in isolation over ashort time period or with a single cross section of data

The estimates from our main specications suggest that theEITC and other tax changes account for over 60 percent of the1984 ndash1996 increase in the weekly and annual employment ofsingle mothers (relative to single women without children) Wel-fare waivers and other changes in AFDC account for smaller butstill large shares of the increase for both employment measuresChanges in Medicaid training and child care programs play asmaller role Our estimated effects of tax and EITC changes arefairly robust across time periods and specications We ndlarger effects for less educated women and smaller but stillsubstantial effects when we compare changes for single motherswith different numbers of children Some of these identicationstrategies result in much weaker AFDC effects The effects ofother policies on employment tend not to vary much by specica-tion Additionally we nd that the effects of the policies on totalhours worked are very similar to the employment results

The structure of the paper is as follows Section II provides atheory of the decision to work and states our main modelingchoices We describe the two data sets used in the empirical workin Section III Section IV describes the main program changes

5 See Gueron and Pauly [1991] for a review of training programs for welfarerecipients and Council of Economic Advisers [1997] for a review of work on theeffects of child care

6 See Levine and Whitmore [1998] Martini and Wiseman [1997] Blank[1997] and Ziliak et al [1997] for differing views of the relative importance ofwelfare waivers economic conditions and benet cuts in the recent decline inwelfare receipt

1066 QUARTERLY JOURNAL OF ECONOMICS

over the 1984 to 1996 period that affected the labor supply ofsingle mothers We also provide summary statistics on programchanges discuss their theoretical impacts on labor supply andintroduce variables that measure particular policies Section Vcompares the employment rates and other characteristics of sin-gle mothers and single women without children Section VI in-vestigates how the employment of single mothers was affected bythe policies we study We also examine alternative explanationsfor our results and briey examine hours worked Section VIIprovides an accounting of the contribution of different policychanges to the overall increase in employment of single mothersin recent years We then offer conclusions in Section VIII

II MODELING THE WORK DECISION

Our modeling approach combines some of the best aspects ofstructural methods and quasi-experimental or natural experi-ment type approaches Beginning from a structural approachclaries which variables should enter the work decision and theform in which they should enter Our simple structural modelalso allows us to test some fundamental economic predictions andmore convincingly simulate policy changes7 The quasi-experi-mental methods make transparent the assumptions that allowthe identication of our key coefcients By the appropriate use ofcontrol variables and simplifying assumptions we identify ourkey parameters using only the sources of variation in our ex-planatory variables that we believe are exogenous

We focus on employment because previous work has foundthat women are more responsive to wages and income in thedecision to work than in the hours decision (see Heckman [1993])The probability that a single woman works is just the probabilitythat the expected utility when working Uw exceeds the expectedutility when not working Un w ie Pr[Uw gt Unw ] We takeutility to be a function of income Y nonmarket time L an indi-cator for welfare participation P (which captures transactioncosts or stigma) other demographic and other control variablesX and an additive stochastic term e Thus the probability ofwork is just

7 Because of the simplications we make to improve the modelrsquos tractabilityone may not want to consider our approach fully structural As with any struc-tural model simulations that rely heavily on simplifying assumptions may givemisleading results

1067THE LABOR SUPPLY OF SINGLE MOTHERS

(1) Pr[U(Yw Lw Pw X) U(Ynw Lnw Pnw X)]where the randomness in this event comes from the stochasticterm e

Income when working is pretax earnings minus taxes plusAFDC and Food Stamps plus Medicaid benets Income whennot working is the maximum AFDCFood Stamp benet andMedicaid benets In each case we calculate the earnings taxesand benets for a given individual incorporating family composi-tion (number and ages of children) and characteristics of stateand federal policies at the time We calculate real income andbenets across states using a cost of living index that depends onstate housing costs The decision to work should depend on thereal return to work not the nominal return8

A key issue in implementing this approach is the form of theuncertainty about a womanrsquos wage and hours should she work Inthe estimates reported here we take a woman to have no moreknowledge of her potential wage and hours than we do as re-searchers9 Thus we take her wage to be a random draw from adistribution (to be specied below) and her hours worked to be arandom draw from a distribution (also to be specied below) thatis conditional on the wage realization Then the probability ofworking is just

(2) PrE[Uw] Unwwhere the expectation here is over the joint wage and hoursdistribution

To estimate equation (2) we take the distribution of e to benormal and take U to be linear in income and nonmarket time (wehave relaxed this latter assumption in other work) In the linearcase (2) has a very simple form

(3) Pra (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw)

2 r (E[Pw] 2 Pnw) 1 X 9 g e nw 2 e wwhere X is other variables that may affect the work decision such

8 Our base specication includes a state cost of living adjustment followingthe approach of National Research Council [1995] One can argue that housingcosts largely reect local amenities However to the extent that these amenitiesare largely xed benets of an area one would still want to account for statedifferences in housing costs when calculating the value of additional income

9 We have also considered two alternatives 1) a woman knows her wage andhours before choosing to work and 2) a woman knows her wage but not her hoursbefore choosing to work Our experiments with these alternatives yielded resultsqualitatively similar to our main results

1068 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

The largest change in the work incentives of single mothersbetween 1984 and 1996 was a tenfold increase in credits throughthe Earned Income Tax Credit (EITC) Only working families(primarily those with children) receive the EITC so its expansionincreased the incentive for single mothers to work We analyzenot only the federal EITC but other federal income tax changesstate income taxes and state EITCs as well The Medicaid pro-gram also greatly expanded during this period Between 1984 and1994 the number of children receiving health coverage throughMedicaid increased 77 percent while the number of coveredadults with dependent children increased 35 percent The expan-sions increased coverage for nonwelfare families with incomesnear the poverty line thus making work more attractive forlow-income single mothers Cash assistance to single parentsthrough Aid to Families with Dependent Children (AFDC) alsochanged quite dramatically over this period Nearly every stateexperimented with changes often under federal waivers of theexisting AFDC rules These changes typically imposed work re-quirements time limits or other measures to encourage singlemothers to work We also investigate the effects of other changesto the AFDC and Food Stamp programs including changes inbenet levels earnings disregards and benet reduction ratesFinally we examine the effects of changes in child care andtraining programs during this period

Our main research strategy identies the effects of thesepolicies on single mothersrsquo labor supply through the differentialtreatment of single mothers and single women without childrenunder welfare and tax laws However the richness of these policychanges allows us to consider additional specications that focuson narrower sources of variation including differences amongsingle mothers in their numbers and ages of children and differ-ences across states in their taxes benets and living costs Thesesources of variation are likely to be unrelated to underlyingdifferences across individuals in their desire to work and thusare likely to be exogenous to labor supply decisions We alsodevelop a new methodology for summarizing the key features ofthe complex nonlinear budget sets created by policies such as theEITC Medicaid and AFDC

Understanding the relationship between the changes in gov-ernment policies and the increases in the labor supply of singlemothers during this period is important for several reasons Firstthese changes in policies provide a plausible source of exogenous

1064 QUARTERLY JOURNAL OF ECONOMICS

variation with which to identify the effects of tax and welfareparameters on labor supply The magnitudes of these effects arekey determinants of the gains or losses from changes in incomeredistribution and social insurance policies

Second understanding the effects of government policiesduring the 1984 ndash1996 period has taken on more importance dueto the passage of the Personal Responsibility and Work Opportu-nity Reconciliation Act of 1996 (PRWORA) In 1997 PRWORAreplaced the main cash assistance program for single mothersAFDC with Temporary Assistance for Needy Families (TANF)The increased state discretion under the new law combined withpolitical changes has led to welfare reform which discourageswelfare receipt and often diverts potential welfare recipients fromtraditional programs These reforms are difcult if not impossibleto characterize using a few variables It is likely that many of thepolicies examined in this paper will be harder and more problem-atic to analyze using post-PRWORA data1

Third there is surprisingly little previous work that esti-mates the effects of the EITC Medicaid or welfare changes onwhether single mothers work The only paper that directly exam-ines how the EITC affects single mothersrsquo labor supply is Eissaand Liebman [1996] which examines the effect of the Tax ReformAct of 19862 In his discussion of the labor supply effects ofMedicaid Moftt [1992] argues that there has been too little workto draw reliable conclusions3 Moftt describes the labor supplyeffect of AFDC as being subject to considerable uncertainty andnotes that the broader labor supply literature has examinedsingle mothers ldquoonly rarelyrdquo4 Dickert Houser and Scholz [1995]argue that this literature provides little guidance as to how theEITC will affect labor market participation and that this omis-sion is especially important because past work suggests that most

1 See Ellwood [2000] National Research Council [1999] and Jencks andSwingle [2000] for related arguments

2 Several papers use labor supply parameters estimated from the negativeincome tax experiments and other sources to simulate the effects of the EITCincluding Hoffman and Seidman [1990] Holtzblatt McCubbin and Gillette[1994] Browning [1995] and Dickert Houser and Scholz [1995] Dickert Houserand Scholz estimate the effect of the after-tax wage and welfare programs onparticipation using a cross section of data from the 1990 panel of the Survey ofIncome and Program Participation (SIPP) They then apply these results tosimulate the effects of the EITC on participation Eissa and Hoynes [1998]examine the effects of the EITC on the labor supply of married couples

3 See Blank [1989] Winkler [1991] and Moftt and Wolfe [1992] in particu-lar The more recent work of Yelowitz [1995] examines the 1988 to 1991 period

4 See Danziger Haveman and Plotnick [1981] and Moftt [1992]

1065THE LABOR SUPPLY OF SINGLE MOTHERS

of the labor supply response is in the work decision rather thanthe hours decision Furthermore there is no work that we areaware of that assesses the overall effect of recent changes intraining and child care programs5 The work on the effects ofwelfare waivers has examined program caseloads rather thanemployment and has reached conicting results6

We examine the major policies affecting the labor supply ofsingle mothers during the 1984 to 1996 period using two datasets the Current Population Survey (CPS) Outgoing RotationGroup Files and the March CPS Files By investigating severalprograms at once using thirteen years of individual data weaccount for their separate effects and we can directly comparethe programs using the same sample time period and methodsOur approach improves on the common past research strategy ofexamining changes in one of these policies in isolation over ashort time period or with a single cross section of data

The estimates from our main specications suggest that theEITC and other tax changes account for over 60 percent of the1984 ndash1996 increase in the weekly and annual employment ofsingle mothers (relative to single women without children) Wel-fare waivers and other changes in AFDC account for smaller butstill large shares of the increase for both employment measuresChanges in Medicaid training and child care programs play asmaller role Our estimated effects of tax and EITC changes arefairly robust across time periods and specications We ndlarger effects for less educated women and smaller but stillsubstantial effects when we compare changes for single motherswith different numbers of children Some of these identicationstrategies result in much weaker AFDC effects The effects ofother policies on employment tend not to vary much by specica-tion Additionally we nd that the effects of the policies on totalhours worked are very similar to the employment results

The structure of the paper is as follows Section II provides atheory of the decision to work and states our main modelingchoices We describe the two data sets used in the empirical workin Section III Section IV describes the main program changes

5 See Gueron and Pauly [1991] for a review of training programs for welfarerecipients and Council of Economic Advisers [1997] for a review of work on theeffects of child care

6 See Levine and Whitmore [1998] Martini and Wiseman [1997] Blank[1997] and Ziliak et al [1997] for differing views of the relative importance ofwelfare waivers economic conditions and benet cuts in the recent decline inwelfare receipt

1066 QUARTERLY JOURNAL OF ECONOMICS

over the 1984 to 1996 period that affected the labor supply ofsingle mothers We also provide summary statistics on programchanges discuss their theoretical impacts on labor supply andintroduce variables that measure particular policies Section Vcompares the employment rates and other characteristics of sin-gle mothers and single women without children Section VI in-vestigates how the employment of single mothers was affected bythe policies we study We also examine alternative explanationsfor our results and briey examine hours worked Section VIIprovides an accounting of the contribution of different policychanges to the overall increase in employment of single mothersin recent years We then offer conclusions in Section VIII

II MODELING THE WORK DECISION

Our modeling approach combines some of the best aspects ofstructural methods and quasi-experimental or natural experi-ment type approaches Beginning from a structural approachclaries which variables should enter the work decision and theform in which they should enter Our simple structural modelalso allows us to test some fundamental economic predictions andmore convincingly simulate policy changes7 The quasi-experi-mental methods make transparent the assumptions that allowthe identication of our key coefcients By the appropriate use ofcontrol variables and simplifying assumptions we identify ourkey parameters using only the sources of variation in our ex-planatory variables that we believe are exogenous

We focus on employment because previous work has foundthat women are more responsive to wages and income in thedecision to work than in the hours decision (see Heckman [1993])The probability that a single woman works is just the probabilitythat the expected utility when working Uw exceeds the expectedutility when not working Un w ie Pr[Uw gt Unw ] We takeutility to be a function of income Y nonmarket time L an indi-cator for welfare participation P (which captures transactioncosts or stigma) other demographic and other control variablesX and an additive stochastic term e Thus the probability ofwork is just

7 Because of the simplications we make to improve the modelrsquos tractabilityone may not want to consider our approach fully structural As with any struc-tural model simulations that rely heavily on simplifying assumptions may givemisleading results

1067THE LABOR SUPPLY OF SINGLE MOTHERS

(1) Pr[U(Yw Lw Pw X) U(Ynw Lnw Pnw X)]where the randomness in this event comes from the stochasticterm e

Income when working is pretax earnings minus taxes plusAFDC and Food Stamps plus Medicaid benets Income whennot working is the maximum AFDCFood Stamp benet andMedicaid benets In each case we calculate the earnings taxesand benets for a given individual incorporating family composi-tion (number and ages of children) and characteristics of stateand federal policies at the time We calculate real income andbenets across states using a cost of living index that depends onstate housing costs The decision to work should depend on thereal return to work not the nominal return8

A key issue in implementing this approach is the form of theuncertainty about a womanrsquos wage and hours should she work Inthe estimates reported here we take a woman to have no moreknowledge of her potential wage and hours than we do as re-searchers9 Thus we take her wage to be a random draw from adistribution (to be specied below) and her hours worked to be arandom draw from a distribution (also to be specied below) thatis conditional on the wage realization Then the probability ofworking is just

(2) PrE[Uw] Unwwhere the expectation here is over the joint wage and hoursdistribution

To estimate equation (2) we take the distribution of e to benormal and take U to be linear in income and nonmarket time (wehave relaxed this latter assumption in other work) In the linearcase (2) has a very simple form

(3) Pra (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw)

2 r (E[Pw] 2 Pnw) 1 X 9 g e nw 2 e wwhere X is other variables that may affect the work decision such

8 Our base specication includes a state cost of living adjustment followingthe approach of National Research Council [1995] One can argue that housingcosts largely reect local amenities However to the extent that these amenitiesare largely xed benets of an area one would still want to account for statedifferences in housing costs when calculating the value of additional income

9 We have also considered two alternatives 1) a woman knows her wage andhours before choosing to work and 2) a woman knows her wage but not her hoursbefore choosing to work Our experiments with these alternatives yielded resultsqualitatively similar to our main results

1068 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

variation with which to identify the effects of tax and welfareparameters on labor supply The magnitudes of these effects arekey determinants of the gains or losses from changes in incomeredistribution and social insurance policies

Second understanding the effects of government policiesduring the 1984 ndash1996 period has taken on more importance dueto the passage of the Personal Responsibility and Work Opportu-nity Reconciliation Act of 1996 (PRWORA) In 1997 PRWORAreplaced the main cash assistance program for single mothersAFDC with Temporary Assistance for Needy Families (TANF)The increased state discretion under the new law combined withpolitical changes has led to welfare reform which discourageswelfare receipt and often diverts potential welfare recipients fromtraditional programs These reforms are difcult if not impossibleto characterize using a few variables It is likely that many of thepolicies examined in this paper will be harder and more problem-atic to analyze using post-PRWORA data1

Third there is surprisingly little previous work that esti-mates the effects of the EITC Medicaid or welfare changes onwhether single mothers work The only paper that directly exam-ines how the EITC affects single mothersrsquo labor supply is Eissaand Liebman [1996] which examines the effect of the Tax ReformAct of 19862 In his discussion of the labor supply effects ofMedicaid Moftt [1992] argues that there has been too little workto draw reliable conclusions3 Moftt describes the labor supplyeffect of AFDC as being subject to considerable uncertainty andnotes that the broader labor supply literature has examinedsingle mothers ldquoonly rarelyrdquo4 Dickert Houser and Scholz [1995]argue that this literature provides little guidance as to how theEITC will affect labor market participation and that this omis-sion is especially important because past work suggests that most

1 See Ellwood [2000] National Research Council [1999] and Jencks andSwingle [2000] for related arguments

2 Several papers use labor supply parameters estimated from the negativeincome tax experiments and other sources to simulate the effects of the EITCincluding Hoffman and Seidman [1990] Holtzblatt McCubbin and Gillette[1994] Browning [1995] and Dickert Houser and Scholz [1995] Dickert Houserand Scholz estimate the effect of the after-tax wage and welfare programs onparticipation using a cross section of data from the 1990 panel of the Survey ofIncome and Program Participation (SIPP) They then apply these results tosimulate the effects of the EITC on participation Eissa and Hoynes [1998]examine the effects of the EITC on the labor supply of married couples

3 See Blank [1989] Winkler [1991] and Moftt and Wolfe [1992] in particu-lar The more recent work of Yelowitz [1995] examines the 1988 to 1991 period

4 See Danziger Haveman and Plotnick [1981] and Moftt [1992]

1065THE LABOR SUPPLY OF SINGLE MOTHERS

of the labor supply response is in the work decision rather thanthe hours decision Furthermore there is no work that we areaware of that assesses the overall effect of recent changes intraining and child care programs5 The work on the effects ofwelfare waivers has examined program caseloads rather thanemployment and has reached conicting results6

We examine the major policies affecting the labor supply ofsingle mothers during the 1984 to 1996 period using two datasets the Current Population Survey (CPS) Outgoing RotationGroup Files and the March CPS Files By investigating severalprograms at once using thirteen years of individual data weaccount for their separate effects and we can directly comparethe programs using the same sample time period and methodsOur approach improves on the common past research strategy ofexamining changes in one of these policies in isolation over ashort time period or with a single cross section of data

The estimates from our main specications suggest that theEITC and other tax changes account for over 60 percent of the1984 ndash1996 increase in the weekly and annual employment ofsingle mothers (relative to single women without children) Wel-fare waivers and other changes in AFDC account for smaller butstill large shares of the increase for both employment measuresChanges in Medicaid training and child care programs play asmaller role Our estimated effects of tax and EITC changes arefairly robust across time periods and specications We ndlarger effects for less educated women and smaller but stillsubstantial effects when we compare changes for single motherswith different numbers of children Some of these identicationstrategies result in much weaker AFDC effects The effects ofother policies on employment tend not to vary much by specica-tion Additionally we nd that the effects of the policies on totalhours worked are very similar to the employment results

The structure of the paper is as follows Section II provides atheory of the decision to work and states our main modelingchoices We describe the two data sets used in the empirical workin Section III Section IV describes the main program changes

5 See Gueron and Pauly [1991] for a review of training programs for welfarerecipients and Council of Economic Advisers [1997] for a review of work on theeffects of child care

6 See Levine and Whitmore [1998] Martini and Wiseman [1997] Blank[1997] and Ziliak et al [1997] for differing views of the relative importance ofwelfare waivers economic conditions and benet cuts in the recent decline inwelfare receipt

1066 QUARTERLY JOURNAL OF ECONOMICS

over the 1984 to 1996 period that affected the labor supply ofsingle mothers We also provide summary statistics on programchanges discuss their theoretical impacts on labor supply andintroduce variables that measure particular policies Section Vcompares the employment rates and other characteristics of sin-gle mothers and single women without children Section VI in-vestigates how the employment of single mothers was affected bythe policies we study We also examine alternative explanationsfor our results and briey examine hours worked Section VIIprovides an accounting of the contribution of different policychanges to the overall increase in employment of single mothersin recent years We then offer conclusions in Section VIII

II MODELING THE WORK DECISION

Our modeling approach combines some of the best aspects ofstructural methods and quasi-experimental or natural experi-ment type approaches Beginning from a structural approachclaries which variables should enter the work decision and theform in which they should enter Our simple structural modelalso allows us to test some fundamental economic predictions andmore convincingly simulate policy changes7 The quasi-experi-mental methods make transparent the assumptions that allowthe identication of our key coefcients By the appropriate use ofcontrol variables and simplifying assumptions we identify ourkey parameters using only the sources of variation in our ex-planatory variables that we believe are exogenous

We focus on employment because previous work has foundthat women are more responsive to wages and income in thedecision to work than in the hours decision (see Heckman [1993])The probability that a single woman works is just the probabilitythat the expected utility when working Uw exceeds the expectedutility when not working Un w ie Pr[Uw gt Unw ] We takeutility to be a function of income Y nonmarket time L an indi-cator for welfare participation P (which captures transactioncosts or stigma) other demographic and other control variablesX and an additive stochastic term e Thus the probability ofwork is just

7 Because of the simplications we make to improve the modelrsquos tractabilityone may not want to consider our approach fully structural As with any struc-tural model simulations that rely heavily on simplifying assumptions may givemisleading results

1067THE LABOR SUPPLY OF SINGLE MOTHERS

(1) Pr[U(Yw Lw Pw X) U(Ynw Lnw Pnw X)]where the randomness in this event comes from the stochasticterm e

Income when working is pretax earnings minus taxes plusAFDC and Food Stamps plus Medicaid benets Income whennot working is the maximum AFDCFood Stamp benet andMedicaid benets In each case we calculate the earnings taxesand benets for a given individual incorporating family composi-tion (number and ages of children) and characteristics of stateand federal policies at the time We calculate real income andbenets across states using a cost of living index that depends onstate housing costs The decision to work should depend on thereal return to work not the nominal return8

A key issue in implementing this approach is the form of theuncertainty about a womanrsquos wage and hours should she work Inthe estimates reported here we take a woman to have no moreknowledge of her potential wage and hours than we do as re-searchers9 Thus we take her wage to be a random draw from adistribution (to be specied below) and her hours worked to be arandom draw from a distribution (also to be specied below) thatis conditional on the wage realization Then the probability ofworking is just

(2) PrE[Uw] Unwwhere the expectation here is over the joint wage and hoursdistribution

To estimate equation (2) we take the distribution of e to benormal and take U to be linear in income and nonmarket time (wehave relaxed this latter assumption in other work) In the linearcase (2) has a very simple form

(3) Pra (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw)

2 r (E[Pw] 2 Pnw) 1 X 9 g e nw 2 e wwhere X is other variables that may affect the work decision such

8 Our base specication includes a state cost of living adjustment followingthe approach of National Research Council [1995] One can argue that housingcosts largely reect local amenities However to the extent that these amenitiesare largely xed benets of an area one would still want to account for statedifferences in housing costs when calculating the value of additional income

9 We have also considered two alternatives 1) a woman knows her wage andhours before choosing to work and 2) a woman knows her wage but not her hoursbefore choosing to work Our experiments with these alternatives yielded resultsqualitatively similar to our main results

1068 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

of the labor supply response is in the work decision rather thanthe hours decision Furthermore there is no work that we areaware of that assesses the overall effect of recent changes intraining and child care programs5 The work on the effects ofwelfare waivers has examined program caseloads rather thanemployment and has reached conicting results6

We examine the major policies affecting the labor supply ofsingle mothers during the 1984 to 1996 period using two datasets the Current Population Survey (CPS) Outgoing RotationGroup Files and the March CPS Files By investigating severalprograms at once using thirteen years of individual data weaccount for their separate effects and we can directly comparethe programs using the same sample time period and methodsOur approach improves on the common past research strategy ofexamining changes in one of these policies in isolation over ashort time period or with a single cross section of data

The estimates from our main specications suggest that theEITC and other tax changes account for over 60 percent of the1984 ndash1996 increase in the weekly and annual employment ofsingle mothers (relative to single women without children) Wel-fare waivers and other changes in AFDC account for smaller butstill large shares of the increase for both employment measuresChanges in Medicaid training and child care programs play asmaller role Our estimated effects of tax and EITC changes arefairly robust across time periods and specications We ndlarger effects for less educated women and smaller but stillsubstantial effects when we compare changes for single motherswith different numbers of children Some of these identicationstrategies result in much weaker AFDC effects The effects ofother policies on employment tend not to vary much by specica-tion Additionally we nd that the effects of the policies on totalhours worked are very similar to the employment results

The structure of the paper is as follows Section II provides atheory of the decision to work and states our main modelingchoices We describe the two data sets used in the empirical workin Section III Section IV describes the main program changes

5 See Gueron and Pauly [1991] for a review of training programs for welfarerecipients and Council of Economic Advisers [1997] for a review of work on theeffects of child care

6 See Levine and Whitmore [1998] Martini and Wiseman [1997] Blank[1997] and Ziliak et al [1997] for differing views of the relative importance ofwelfare waivers economic conditions and benet cuts in the recent decline inwelfare receipt

1066 QUARTERLY JOURNAL OF ECONOMICS

over the 1984 to 1996 period that affected the labor supply ofsingle mothers We also provide summary statistics on programchanges discuss their theoretical impacts on labor supply andintroduce variables that measure particular policies Section Vcompares the employment rates and other characteristics of sin-gle mothers and single women without children Section VI in-vestigates how the employment of single mothers was affected bythe policies we study We also examine alternative explanationsfor our results and briey examine hours worked Section VIIprovides an accounting of the contribution of different policychanges to the overall increase in employment of single mothersin recent years We then offer conclusions in Section VIII

II MODELING THE WORK DECISION

Our modeling approach combines some of the best aspects ofstructural methods and quasi-experimental or natural experi-ment type approaches Beginning from a structural approachclaries which variables should enter the work decision and theform in which they should enter Our simple structural modelalso allows us to test some fundamental economic predictions andmore convincingly simulate policy changes7 The quasi-experi-mental methods make transparent the assumptions that allowthe identication of our key coefcients By the appropriate use ofcontrol variables and simplifying assumptions we identify ourkey parameters using only the sources of variation in our ex-planatory variables that we believe are exogenous

We focus on employment because previous work has foundthat women are more responsive to wages and income in thedecision to work than in the hours decision (see Heckman [1993])The probability that a single woman works is just the probabilitythat the expected utility when working Uw exceeds the expectedutility when not working Un w ie Pr[Uw gt Unw ] We takeutility to be a function of income Y nonmarket time L an indi-cator for welfare participation P (which captures transactioncosts or stigma) other demographic and other control variablesX and an additive stochastic term e Thus the probability ofwork is just

7 Because of the simplications we make to improve the modelrsquos tractabilityone may not want to consider our approach fully structural As with any struc-tural model simulations that rely heavily on simplifying assumptions may givemisleading results

1067THE LABOR SUPPLY OF SINGLE MOTHERS

(1) Pr[U(Yw Lw Pw X) U(Ynw Lnw Pnw X)]where the randomness in this event comes from the stochasticterm e

Income when working is pretax earnings minus taxes plusAFDC and Food Stamps plus Medicaid benets Income whennot working is the maximum AFDCFood Stamp benet andMedicaid benets In each case we calculate the earnings taxesand benets for a given individual incorporating family composi-tion (number and ages of children) and characteristics of stateand federal policies at the time We calculate real income andbenets across states using a cost of living index that depends onstate housing costs The decision to work should depend on thereal return to work not the nominal return8

A key issue in implementing this approach is the form of theuncertainty about a womanrsquos wage and hours should she work Inthe estimates reported here we take a woman to have no moreknowledge of her potential wage and hours than we do as re-searchers9 Thus we take her wage to be a random draw from adistribution (to be specied below) and her hours worked to be arandom draw from a distribution (also to be specied below) thatis conditional on the wage realization Then the probability ofworking is just

(2) PrE[Uw] Unwwhere the expectation here is over the joint wage and hoursdistribution

To estimate equation (2) we take the distribution of e to benormal and take U to be linear in income and nonmarket time (wehave relaxed this latter assumption in other work) In the linearcase (2) has a very simple form

(3) Pra (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw)

2 r (E[Pw] 2 Pnw) 1 X 9 g e nw 2 e wwhere X is other variables that may affect the work decision such

8 Our base specication includes a state cost of living adjustment followingthe approach of National Research Council [1995] One can argue that housingcosts largely reect local amenities However to the extent that these amenitiesare largely xed benets of an area one would still want to account for statedifferences in housing costs when calculating the value of additional income

9 We have also considered two alternatives 1) a woman knows her wage andhours before choosing to work and 2) a woman knows her wage but not her hoursbefore choosing to work Our experiments with these alternatives yielded resultsqualitatively similar to our main results

1068 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

over the 1984 to 1996 period that affected the labor supply ofsingle mothers We also provide summary statistics on programchanges discuss their theoretical impacts on labor supply andintroduce variables that measure particular policies Section Vcompares the employment rates and other characteristics of sin-gle mothers and single women without children Section VI in-vestigates how the employment of single mothers was affected bythe policies we study We also examine alternative explanationsfor our results and briey examine hours worked Section VIIprovides an accounting of the contribution of different policychanges to the overall increase in employment of single mothersin recent years We then offer conclusions in Section VIII

II MODELING THE WORK DECISION

Our modeling approach combines some of the best aspects ofstructural methods and quasi-experimental or natural experi-ment type approaches Beginning from a structural approachclaries which variables should enter the work decision and theform in which they should enter Our simple structural modelalso allows us to test some fundamental economic predictions andmore convincingly simulate policy changes7 The quasi-experi-mental methods make transparent the assumptions that allowthe identication of our key coefcients By the appropriate use ofcontrol variables and simplifying assumptions we identify ourkey parameters using only the sources of variation in our ex-planatory variables that we believe are exogenous

We focus on employment because previous work has foundthat women are more responsive to wages and income in thedecision to work than in the hours decision (see Heckman [1993])The probability that a single woman works is just the probabilitythat the expected utility when working Uw exceeds the expectedutility when not working Un w ie Pr[Uw gt Unw ] We takeutility to be a function of income Y nonmarket time L an indi-cator for welfare participation P (which captures transactioncosts or stigma) other demographic and other control variablesX and an additive stochastic term e Thus the probability ofwork is just

7 Because of the simplications we make to improve the modelrsquos tractabilityone may not want to consider our approach fully structural As with any struc-tural model simulations that rely heavily on simplifying assumptions may givemisleading results

1067THE LABOR SUPPLY OF SINGLE MOTHERS

(1) Pr[U(Yw Lw Pw X) U(Ynw Lnw Pnw X)]where the randomness in this event comes from the stochasticterm e

Income when working is pretax earnings minus taxes plusAFDC and Food Stamps plus Medicaid benets Income whennot working is the maximum AFDCFood Stamp benet andMedicaid benets In each case we calculate the earnings taxesand benets for a given individual incorporating family composi-tion (number and ages of children) and characteristics of stateand federal policies at the time We calculate real income andbenets across states using a cost of living index that depends onstate housing costs The decision to work should depend on thereal return to work not the nominal return8

A key issue in implementing this approach is the form of theuncertainty about a womanrsquos wage and hours should she work Inthe estimates reported here we take a woman to have no moreknowledge of her potential wage and hours than we do as re-searchers9 Thus we take her wage to be a random draw from adistribution (to be specied below) and her hours worked to be arandom draw from a distribution (also to be specied below) thatis conditional on the wage realization Then the probability ofworking is just

(2) PrE[Uw] Unwwhere the expectation here is over the joint wage and hoursdistribution

To estimate equation (2) we take the distribution of e to benormal and take U to be linear in income and nonmarket time (wehave relaxed this latter assumption in other work) In the linearcase (2) has a very simple form

(3) Pra (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw)

2 r (E[Pw] 2 Pnw) 1 X 9 g e nw 2 e wwhere X is other variables that may affect the work decision such

8 Our base specication includes a state cost of living adjustment followingthe approach of National Research Council [1995] One can argue that housingcosts largely reect local amenities However to the extent that these amenitiesare largely xed benets of an area one would still want to account for statedifferences in housing costs when calculating the value of additional income

9 We have also considered two alternatives 1) a woman knows her wage andhours before choosing to work and 2) a woman knows her wage but not her hoursbefore choosing to work Our experiments with these alternatives yielded resultsqualitatively similar to our main results

1068 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

(1) Pr[U(Yw Lw Pw X) U(Ynw Lnw Pnw X)]where the randomness in this event comes from the stochasticterm e

Income when working is pretax earnings minus taxes plusAFDC and Food Stamps plus Medicaid benets Income whennot working is the maximum AFDCFood Stamp benet andMedicaid benets In each case we calculate the earnings taxesand benets for a given individual incorporating family composi-tion (number and ages of children) and characteristics of stateand federal policies at the time We calculate real income andbenets across states using a cost of living index that depends onstate housing costs The decision to work should depend on thereal return to work not the nominal return8

A key issue in implementing this approach is the form of theuncertainty about a womanrsquos wage and hours should she work Inthe estimates reported here we take a woman to have no moreknowledge of her potential wage and hours than we do as re-searchers9 Thus we take her wage to be a random draw from adistribution (to be specied below) and her hours worked to be arandom draw from a distribution (also to be specied below) thatis conditional on the wage realization Then the probability ofworking is just

(2) PrE[Uw] Unwwhere the expectation here is over the joint wage and hoursdistribution

To estimate equation (2) we take the distribution of e to benormal and take U to be linear in income and nonmarket time (wehave relaxed this latter assumption in other work) In the linearcase (2) has a very simple form

(3) Pra (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw)

2 r (E[Pw] 2 Pnw) 1 X 9 g e nw 2 e wwhere X is other variables that may affect the work decision such

8 Our base specication includes a state cost of living adjustment followingthe approach of National Research Council [1995] One can argue that housingcosts largely reect local amenities However to the extent that these amenitiesare largely xed benets of an area one would still want to account for statedifferences in housing costs when calculating the value of additional income

9 We have also considered two alternatives 1) a woman knows her wage andhours before choosing to work and 2) a woman knows her wage but not her hoursbefore choosing to work Our experiments with these alternatives yielded resultsqualitatively similar to our main results

1068 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

as demographic variables and characteristics of state welfarewaivers training programs and child care programs This speci-cation also allows xed costs of work which vary across demo-graphic groups Under the normality assumption (3) can be re-written as

(4) F a (E[Yw] 2 Ynw) 1 b (E[Lw] 2 Lnw) 2 r (E[Pw] 2 Pnw) 1 X 9 g We make the simplifying assumption that nonworking singlemothers participate in welfare and that working single mothersparticipate if their earnings are low enough to qualify them foraid This assumption is clearly a simplication as some womenwho qualify for aid will not participate because of the transactioncosts or stigma of doing so Past work on program takeup sug-gests that about 75 percent of those eligible for AFDC and about50 percent of those eligible for Food Stamps participate (for arecent review of past work see Blank and Ruggles [1996]) How-ever AFDC takeup rates between 80 and 90 percent are probablycloser to the truth given the underreporting of welfare receipt instandard data sets (see Bavier [1999]) We also assume that allsingle women without children do not participate in welfare pro-grams10

We generalize (4) by allowing the coefcients on the differentcomponents of income to differ since income from differentsources may be valued differently For example we allow theeffect of welfare income (AFDC plus Food Stamps) to differ fromthat of labor income taxes paid and Medicaid coverage Welfareincome may be valued less than labor income because of a vari-able component to the transaction costs or stigma of welfareparticipation (see Moftt [1983]) Medicaid may be valued at lessthan our calculated cost because it is an in-kind transfer or morethan cost because of its insurance component These separatecoefcients on different income terms allow for additional tests ofthe hypothesis that increases in the return to work make workmore likely and they allow an approach that is less restrictiveie less likely to yield biased estimates

We assume that all single mothers face the same pretax wage

10 The primary program for which single women without children would beeligible is Food Stamps Single adults with children are more than ten times aslikely to receive Food Stamps as single adults without children (authorsrsquo calcula-tions and U S Department of Agriculture [1995]) Furthermore since the FoodStamp program has not changed much over time and does not differ much by stateexcept for interactions with AFDC our control variables below (particularly yearand number of children dummies) should account for most of these differences

1069THE LABOR SUPPLY OF SINGLE MOTHERS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

and hours distribution and we make the same assumption forsingle childless women We estimated some specications thatused a wagehours distribution that varied with demographicsalthough these results are not reported here (see Meyer andRosenbaum [1999]) Thus expected earnings if working only varywith the controls and are absorbed by X which includes variablesfor the presence and number of children age education stateyear and other variables described fully below Similarly non-market time when working and not working E[Lw ] and Lnw respectively are taken to be constant or to vary with X and thusare absorbed by X Pn w which identically equals 1 is absorbedinto the constant We then obtain the employment probability

(5) F a 1E[taxes] 1 a 2E[AFDC and Food Stamp benefits if work]

1 a 3E[Medicaid coverage if work valued at cost]

2 a 4maximum AFDCFood Stamp benefit

2 a 5Medicaid coverage if do not work valued at cost

1 r E[Pw] 1 X 9 g

We allow the tax and welfare variables in (5) to vary withyear state and the number and ages of children To implementthis approach we discretize the wage and hours distribution andperform the numerical integration required in (5) allowing thehours distribution to vary with the wage level because of thepronounced dependence between the two distributions To calcu-late the wage and hours distribution we pool 1984 ndash1996 MarchCPS data and estimate one distribution that we use for all yearsWe do this separately for single mothers and single childlesswomen We approximate these distributions using cells denedby 50 intervals of the joint wage and hours distribution (seeAppendix 1 for details) Our approach is both tractable and yetable to capture the fairly complex and highly nonlinear budgetconstraints of low income single mothers These complexities aredescribed in detail in Section IV

III DATA

The data used in this paper come from the Current Popula-tion Survey (CPS) a nationally representative monthly survey ofapproximately 60000 households We use two types of the CPSdata the March CPS Files and the merged Outgoing Rotation

1070 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

Group (ORG) data During each interview household membersare asked whether they worked last week and their hoursworked as well as many other questions In the March inter-views individuals are asked to provide detailed retrospectiveinformation including hours earnings and weeks worked duringthe previous year The ORG les come from all twelve months ofthe year but only include the same person once in a given yearThe March CPS data are from the 1985ndash1997 interviews andtherefore provide information on the years 1984 ndash1996 The ORGdata are from 1984 ndash1996 We limit the sample to single women(widowed divorced and never married) who are between 19 and44 years old and not in school In the March CPS women whowere ill or disabled during the previous year or who had positiveearned income but zero hours of work are also excluded Theresulting samples sizes are 373662 for the ORG and 119019 forthe March CPS

IV THE POLICY CHANGES AND LABOR SUPPLY

In this section we describe the major policy changes between1984 and 1996 that affected the labor supply of single mothersFor each policy or program we rst provide some brief back-ground information and outline the major changes between 1984and 1996 (see Figure I for a time line depicting these changes)Next we describe the policy variables used in the empirical workto summarize the incentive effects of these programs Finally weanalyze the theoretical effects of these changes on labor supplyespecially on the choice of whether or not to work An in-depthdiscussion of the policy changes is in Meyer and Rosenbaum[2000a]

A The EITC and Federal and State Income Taxes

In our period the most important changes in work incentivesfor single mothers probably came from the Earned Income TaxCredit11 EITC credits increased fteenfold from $16 billion in1984 to a projected $251 billion in 1996 Single parents receivedabout two-thirds of these EITC dollars (see U S House of Rep-resentatives Green Book [1996] U S Department of the Trea-sury SOI [1999]) In 1996 a single woman with two children who

11 See Liebman [1998] for a history of the EITC and a survey of many of thekey economic issues

1071THE LABOR SUPPLY OF SINGLE MOTHERS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

earned less than $8890 (the phase-in range) received a 40 percentcredit on dollars earned up to a maximum of $3556 Because thecredit is refundable and a mother of two with those earnings wasnot subject to any federal income tax (due to the standard deduc-tion and personal exemptions) she would have received a check

FIGURE IMajor Tax and Welfare Policy Changes Affecting Low Income Women

1984ndash1997

1072 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

from the IRS for the credit amount With additional earnings upto $11610 the credit amount did not change Additional earningsbeyond $11610 and up to $28495 (the phase-out range) resultedin a reduction in the credit by 2106 percent of the additionalearnings until the credit was reduced to zero This credit sched-ule meant that a woman with two children earning between$5000 and just under $19000 received at least a $2000 credit

The current EITC is the result of several legislative changes(summarized in Figure I) which greatly expanded the EITC after1984 Between its beginning in 1975 and the passage of the TaxReform Act of 1986 (TRA86) the EITC was small and the creditamounts did not keep up with ination Beginning with theTRA86 the EITC was expanded in a number of dimensions Firstcredit rates phase-in ranges and phase-out ranges were in-creased considerably Second in 1991 the credit was expanded toprovide a larger credit for families with two or more children12

The increment to the maximum credit for a second child wassmall through 1993 but beginning in 1994 the difference began torise sharply it rose to $490 in 1994 $1016 in 1995 and $1404 in1996 Third in 1991 the requirements for qualifying childrenwere changed in a way that tended to increase eligibility

The after-tax incomes of single women were affected by otherchanges in federal income taxes during this period such as the1987 increase in the personal exemption and the 1988 increase inthe standard deduction for household heads To illustrate theoverall changes in after-tax incomes we plot in Figure II thedifference in after-tax income (earnings minus federal incometaxes plus the EITC) between a woman with two children and awoman with no children for various pretax earnings levels in1984 1988 1992 and 199613

Figure II illustrates several important aspects of the EITCexpansions First between 1984 and 1988 single mothers of twowith earnings between $10000 and $20000 experienced in-creases in take-home pay (relative to single women without chil-

12 There were other small program changes From 1991 through 1993 therewere small refundable credits for child health insurance premiums and for chil-dren under one Beginning in January 1991 the EITC was not counted as incomein most means-tested programs increasing its value for very low income women

13 Changes over time in this difference were almost entirely due to changesin the taxes paid (or credits received) by single mothers as can be seen in panel 1of Table I The taxes paid by single women without children hardly changedbetween 1984 and 1996 especially for earnings levels between $10000 and$20000

1073THE LABOR SUPPLY OF SINGLE MOTHERS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

dren) that ranged from $500 to $1500 (unless noted all dollaramounts are in 1996 PCE deated dollars) Most of this increasewas due to large increases in both the maximum credit and theearnings level before the credit phase-out began The most strik-ing feature of Figure II is the large 1994 ndash1996 expansions whichdisproportionately affected women with two or more children Forexample the take-home pay difference for women with $7500 ofearnings increased only about $600 between 1984 and 1993 butincreased over $1500 between 1993 and 1996 Unlike the earlierexpansions those since 1993 dramatically increased the take-home-pay difference for very low income women (earnings under$10000) due to large increases in the credit rate and maximumcredit

As well as federal income tax changes we incorporate in thisstudy the effects of state income taxes including state EITCs By1994 seven states had their own EITCs The largest ve of thesestates began their credit during the period we examine All of thestate EITCs were set as a fraction of the federal EITC and thusincreased when it did There were other state income tax changesduring our sample period that reduced taxes for single mothersMore than a dozen states increased their personal exemptionincreased their child credit added a higher standard deductionor added a separate tax schedule for household heads

To summarize these changes in federal and states taxes we

FIGURE IIAfter-Tax Income of a Single Mother with Two Children Minus a Single

Woman without Children 1984 1988 1992 1996All numbers are in 1996 dollars deated with the Personal Consumption Ex-

penditures Deator All women are assumed to have only earned income and totake the standard deduction Single women with children and without childrenare assumed to le as head of household and single respectively After-tax incomeis income after federal taxes or credits

1074 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

calculate a variable called Income Taxes if Work This variable isthe expected taxes a woman would pay in a given state and yearwith a given family composition and ages of children The expec-tation is calculated by integrating over the wage and hours dis-tribution of single women as described in Section II Appendix 2reports the mean of this variable for single mothers and singlewomen without children for various years Over the years 1984 ndash1996 taxes paid by single mothers relative to single women with-out children fell by $1607 Thirty-nine percent of the relative fallin taxes (increase in credits) occurred in the last three years(1993ndash1996) About 43 percent occurred in 1987 and 1988 with18 percent occurring between 1991 and 1993 Almost all of the fallin relative taxes was due to federal tax changes Only $37 wasdue to state taxes with all but $7 of this due to state EITCsHowever in the seven states with state EITCs the role of statetaxes was much greater In these jurisdictions state EITCs ac-counted for a $215 drop in the taxes of single mothers relative tosingle women without children

The theoretical effect of the EITC expansions on the annualparticipation decision of single parents is unambiguously posi-tive Since the EITC expansions have increased the after-taxreturn to work at all earnings levels work is unambiguouslymore attractive The effect of the EITC and its expansions on thehours of work among those working is much less clear and de-pends on where a person would choose to work on the pre- andpostcredit budget sets Overall the income effect of the creditcombined with the negative substitution effect that people face onthe phase-out portion of the credit is expected to reduce the hoursof those who work14

B AFDC Food Stamps and Waivers

The two programs that have been most commonly thought ofas welfare are Aid to Families with Dependent Children (AFDC)and Food Stamps We discuss Food Stamps along with AFDC

14 One might wonder whether households are aware of these tax incentivesand bother to le tax returns Awareness appears to be high [Romich and Weisner2000 Smeeding et al 2000] and EITC takeup appears to be high and risingScholz [1990 1994] estimates takeup to be 75 percent in 1988 and between 80 and86 percent in 1990 With the increases in the EITC after 1990 that raised thevalue of ling and disproportionately made eligible moderate income people whoare likely to le one might expect that the participation rate rose further Inaddition EITC awareness and outreach has increased in recent years On theother hand recent compliance efforts may have discouraged some potential lers

1075THE LABOR SUPPLY OF SINGLE MOTHERS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

because nearly 90 percent of AFDC recipients also received FoodStamps [U S House of Representatives 1996] The AFDC pro-gram provided cash payments to families with children who havebeen deprived of support due to the absence or unemployment ofa parent The Food Stamp program provides low-income house-holds with coupons to purchase food AFDC program parameterswere set by the states while most Food Stamp parameters are thesame in all states Nevertheless because of the interaction of theeligibility and benet calculations of the two programs there areinterstate differences in the Food Stamps received for people insimilar situations Both of these programs are large relative toother means-tested programs with 1996 AFDC and Food Stampexpenditures totaling $237 billion and $255 billion respectivelyBoth had growing expenditures and caseloads in the late 1980sand early 1990s with peaks in scal year 1994

While much past work has summarized the AFDC and FoodStamp programs using the combined maximum benet this mea-sure ignores the large interstate differences and changes overtime in earnings exemptions and implicit tax rates By 1996fteen states had exemptions and tax rates that differed from thestandard $120 earnings exemption and the two-thirds implicittax rate We summarize AFDC and Food Stamps with threevariables implied by our theoretical model the maximum com-bined benet expected benets if a person works and the prob-ability of AFDC receipt (which captures transaction costs orstigma) Due to cuts in AFDC the mean maximum combinedAFDC and Food Stamp benet fell about 7 percent over thesample period Over the same period mean benets for a workingsingle mother remained roughly constant as implicit tax rateswere reduced

Theory predicts that the AFDC and Food Stamp programsdecrease labor supply for two reasons First the income effect ofthe guarantee amount (maximum benet) should make employ-ment less likely and reduce hours worked if a woman worksSecond the implicit tax rate resulting from reductions in benetsas earnings increase (captured by reductions in the benets ifwork variable) also reduces the incentive to work Thus AFDCshould decrease both the likelihood of working and hours condi-tional on working However in interpreting our estimates belowone should bear in mind that substantial research indicates thatactual exemptions and implicit tax rates differ from the statutory

1076 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

ones15 Consequently our calculations of AFDC benets for thosewho work may be fairly rough We will return to this issue inSection VI

Under AFDC the Secretary of Health and Human Services(HHS) was authorized to waive specied program requirementsto allow states to experiment This waiver authority was rarelyused prior to the late 1980s but its use accelerated under Presi-dents Bush and Clinton Between January 1993 and August1996 HHS approved welfare waivers in 43 states While statesexperimented with changes in nearly every aspect of AFDCmany provisions applied to small parts of states or would not beexpected to have a substantial effect on the employment of singlemothers We focus on a few types of waiver provisions that weretried in many states Our main welfare waiver variables are AnyTime Limit which equals one for single mothers in states thatimposed work requirements or benet reductions on those whoreached time limits and Any Terminations which equals one forany single mother in a state in which a welfare case had beenterminated under a welfare waiver Some common types of pro-visions such as expanded income disregards have been incorpo-rated in our coding of the AFDC program Others such as familycaps (which limited the benets for additional children) or in-creased resource limits (which loosened the asset restrictions forAFDC eligibility) likely have small or ambiguous effects on em-ployment and are therefore not included

In this paper we focus on implementation dates and actualbeginning dates of terminations instead of application or ap-proval dates We also examine a dummy variable for states thatapplied for a major statewide waiver in case this indicates atightening of administrative requirements in a state These vari-ables are interacted with an indicator for whether a woman haschildren In Table I we report the fraction of single women livingin states that have applied for or implemented various types ofwaivers Very few women were in states that had implementedsignicant waivers through at least 1994 The fraction of womenin states that had made a major waiver application was muchhigher 022 in 1992 and 085 in 1996

15 See Fraker Moftt and Wolf [1985] and Levine [1997] Other researchindicates that few AFDC recipients report their income to welfare ofces [Edinand Lein 1997 Hill et al 1999]

1077THE LABOR SUPPLY OF SINGLE MOTHERS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

C Medicaid

Medicaid is the biggest and most costly program that aidssingle mothers and their children In 1994 $309 billion wasspent on 248 million nonaged nondisabled Medicaid recipients agroup that was predominantly single mothers and their children[U S House of Representatives Green Book 1996 pp 897ndash902]Unlike the Food Stamp program and especially AFDC Medicaideligibility has expanded dramatically since 1984 resulting in amore than threefold increase between 1984 and 1994 in Medicaidexpenditures on families with dependent children (and a 60 per-cent increase in the caseload) Prior to 1987 Medicaid eligibilityfor single mothers and their children generally required receipt ofAFDC In a series of expansions Medicaid coverage was extendedto low-income pregnant women and children (again see Figure I)The differences across states in the extent to which they tookadvantage of the permitted coverage options generated largedifferences in who was covered in different years in differentstates Moreover state AFDC income limits interacted with theMedicaid expansions to determine the additional families covered(see Meyer and Rosenbaum [2000b] for more details)

We measure Medicaid benets by rst calculating the num-ber of adults and children in the family that would be covered ifa woman works We then convert these numbers to dollar valuesusing Medicaid expenditures per child and adult averaged overall states and years16 As can be seen in Table I there was a fairlysteady increase over our sample period in the number of familymembers covered under Medicaid if a single mother works

The theoretical effect of Medicaid expansions on the decisionto work is positive since those newly covered are those withearnings that would make them ineligible for AFDC The Med-icaid expansions also could result in some working women in-creasing their hours if pre-expansion earnings limits resulted inthem reducing their hours of work in order to qualify for Medicaidcoverage Overall the effect on hours conditional on working isambiguous since the expansions also could result in hours de-creases for women who choose to reduce their hours in order toqualify for Medicaid coverage for their children

16 Note that in our specications Medicaid coverage for the nonworking iscollinear with family size and number of children controls so a 5 is not estimated

1078 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

D Training and Child Care Programs

To capture the effect of training programs on the probabilityof work by single mothers we focus on the programs specicallyfor AFDC applicants and recipients rst the Work Incentives(WIN) program and then the Job Opportunities and Basic Skills(JOBS) program Total expenditures as well as the emphasis ofthese programs changed sharply over our period (see Table I) Weconstruct two variables that measure the character and extent ofthe JOBS and WIN programs in a state and year Because edu-cational spending is likely to have a different effect than otherspending we split expenditures into education and job searchother We scale state expenditures by the size of the AFDCmandatory population These variables are interacted with anindicator for whether a woman would be required to participate inJOBS or WIN (based on the age of her youngest child these rulesdiffered across states and over time) so that these variables equalzero for single women without children or with children under theage cutoff

The effects of these training programs on labor supply likelydepends on the mix of services provided and the stringency of theparticipation requirements Job search assistance job place-ments and improving job skills and readiness should lower jobsearch costs thereby increasing the level of work for womentrainees On the other hand even with a benecial long-termeffect on wages or employment secondary or postsecondary edu-cation may delay entry into the workforce while women takeclasses leading to a short-term negative employment effect Inany case there is much stronger evidence of employment effectsfrom job search assistance than from education at least in theshort run17

The cost and quality of child care is likely to have an impor-tant effect on whether a woman works The federal role in childcare for low-income women expanded greatly following the Fam-ily Support Act of 1988 and the Omnibus Budget ReconciliationAct of 1990 Four large programs started during this periodAFDC Child Care Transitional Child Care At-Risk Child Careand Child Care and Development Block Grants We focus onthese programs because they are particularly important for singlemothers and they were the main changes over our period Total

17 See Gueron and Pauly [1991] and U S Department of Health andHuman Services [1997b]

1079THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

SU

MM

AR

YC

HA

RA

CT

ER

IST

ICS

OF

PO

LIC

IES

AF

FE

CT

ING

SIN

GL

EM

OT

HE

RS

AN

DS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

841

988

1992

A

ND

1996

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

An

nu

alfe

dera

lst

ate

inco

me

taxe

sE

ITC

an

d1

2O

AS

DH

IA

t$5

000

earn

ings

216

935

22

338

376

253

340

82

1478

194

At

$10

000

earn

ings

118

954

234

713

562

673

1427

220

1214

32A

t$1

500

0ea

rnin

gs15

9920

7578

425

8950

226

872

476

2706

At

$20

000

earn

ings

2721

3325

2477

3844

2374

3980

1686

4009

At

$30

000

earn

ings

5466

6326

5398

6538

5527

6666

5585

6668

An

nu

alA

FD

Can

dfo

odst

amp

bene

ts

At

$0ea

rnin

gs75

830

7406

073

910

7056

0A

t$5

000

earn

ings

4719

047

340

4791

045

640

At

$10

000

earn

ings

1871

018

850

2029

019

750

At

$15

000

earn

ings

491

048

50

640

062

10

At

$20

000

earn

ings

800

890

116

013

20

Med

icai

dn

um

ber

offa

mil

ym

embe

rsel

igib

leA

t$0

earn

ings

265

000

262

000

266

000

268

000

At

$500

0ea

rnin

gs2

510

002

530

002

560

002

520

00A

t$1

000

0ea

rnin

gs1

100

001

410

001

620

001

920

00A

t$1

500

0ea

rnin

gs0

270

000

350

000

760

001

010

00A

t$2

000

0ea

rnin

gs0

030

000

050

000

310

000

490

00A

t$2

500

0ea

rnin

gs0

000

000

010

000

090

000

190

00

1080 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EI

(CO

NT

INU

ED

)

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Wai

vers

An

yti

me

lim

it0

000

000

00

010

039

0A

ny

term

inat

ions

000

00

000

000

00

210

Maj

orw

aive

rap

plic

atio

n0

000

002

00

220

085

0A

nn

ual

trai

nin

gch

ild

care

doll

ars

per

elig

ible

reci

pien

tT

rain

ingmdash

edu

cati

on0

00

010

00

126

0T

rain

ingmdash

job

sear

cho

ther

126

039

016

60

272

0C

hil

dca

re0

00

024

60

302

0N

um

ber

ofob

serv

atio

ns

9391

189

1492

1118

612

103

3319

311

8788

158

46

Sou

rce

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opu

lati

onS

urve

yO

utg

oing

Rot

atio

nG

roup

File

(OR

G)

Res

tric

tion

sT

hesa

mpl

ein

clud

es19

ndash44

year

-old

sing

lew

omen

(div

orce

dw

idow

ed

orne

ver

mar

ried

)w

hoar

eno

tin

scho

ol

Not

esT

hese

mea

nsar

eca

lcu

late

du

sing

the

char

acte

rist

ics

ofth

eO

RG

sam

ple

for

the

give

nye

aran

dar

ew

eigh

ted

Wom

enar

eas

sum

edto

bein

thei

rr

stfo

urm

onth

sof

wor

kto

hav

eno

une

arn

edin

com

ean

dto

clai

mn

och

ildca

reex

pen

ses

Als

osi

ngle

wom

enw

ith

and

wit

hout

chil

dren

are

assu

med

tol

eas

head

ofh

ouse

hol

dan

dsi

ngle

res

pect

ivel

yan

dto

clai

mth

est

anda

rdde

duct

ion

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

san

dal

ldol

lar

amou

nts

are

expr

esse

din

1996

dolla

rsS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1081THE LABOR SUPPLY OF SINGLE MOTHERS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

state and federal expenditures on these four new federal pro-grams by state and year are scaled by the number of singlemothers with children under six These numbers can be seen inTable I which shows a steep rise in child care expendituresbetween 1988 and 1992 followed by a slower rise in later yearsFor more detail on training and child care programs see Meyerand Rosenbaum [1999]

V THE DETERMINANTS OF EMPLOYMENT

We use several different econometric methods to identify theimpact of the recent policy changes on the employment of singlemothers We begin with the familiar difference in differencesestimator This approach compares employment rates over timefor single mothers with those for single women without childrenThis approach is the one taken by Eissa and Liebman [1996] intheir study of the EITC over the 1984 to 1990 period We waituntil Section VI to discuss the estimates from our simple struc-tural model

A Employment Rates of Single Mothers and Single ChildlessWomen

The top panel of Table II reports the employment rates ofsingle mothers and single women without children along withthe difference in employment rates between these two groups ofsingle women We report this difference because many determi-nants of employment that change over time especially wages andmacroeconomic conditions might be expected to affect all singlewomen similarly Other determinants of employment particu-larly the tax and transfer programs that we examine specicallyaffect single mothers The bottom panel of Table II focuses on thesubsample of single mothers with children under six (again rela-tive to single women without children) a group we expect to bemore responsive to changes in the rewards to work Also employ-ment changes are likely to have greater effects on children forbetter or worse when they are young and their mother likelyplays a larger role in their care and education

We report two different measures of employment whether awoman worked last week (from the ORG data) and whether awoman worked at all last year (from the March data) Eachmeasure has its advantages Whether a woman worked last weekis probably a better measure of labor supply to use as an input to

1082 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

policy decisions since its average captures the fraction of womenworking in a given week This variable will be especially useful ifthose who move in or out of the workforce on the margin workonly a few weeks during the year On the other hand as discussedearlier the EITC unequivocally increases the probability of work-

TABLE IIEMPLOYMENT RATES FOR SINGLE MOTHERS SINGLE MOTHERS WITH CHILDREN

UNDER SIX AND SINGLE WOMEN WITHOUT CHILDREN 1984ndash1996

Year

CPS Outgoing Rotation Groupworked last week = 1 March CPS worked last year = 1

ChildrenNo

children DifferenceStandard

error ChildrenNo

children DifferenceStandard

error

1984 05854 08014 2 02160 00059 07322 09399 2 02077 000831985 05861 08048 2 02187 00058 07302 09439 2 02137 000831986 05891 08131 2 02240 00057 07310 09450 2 02141 000821987 05941 08179 2 02238 00056 07382 09473 2 02091 000811988 06027 08215 2 02188 00058 07482 09485 2 02003 000841989 06136 08150 2 02015 00058 07577 09409 2 01831 000801990 06007 08155 2 02148 00056 07591 09424 2 01832 000791991 05790 08031 2 02242 00056 07428 09418 2 01990 000791992 05790 07957 2 02167 00057 07387 09299 2 01913 000811993 05875 07918 2 02044 00057 07511 09356 2 01845 000801994 06053 07921 2 01868 00057 07907 09312 2 01405 000781995 06265 07971 2 01707 00058 08072 09340 2 01268 000801996 06450 07938 2 01488 00060 08191 09290 2 01098 00079

Childrenunder 6

Nochildren Difference

Standarderror

Childrenunder 6

Nochildren Difference

Standarderror

1984 04382 08014 2 03632 00083 06122 09399 2 03277 001311985 04328 08048 2 03720 00082 05966 09439 2 03474 001331986 04362 08131 2 03770 00081 06227 09450 2 03223 001281987 04437 08179 2 03742 00082 06096 09473 2 03377 001291988 04634 08215 2 03581 00084 06277 09485 2 03207 001321989 04790 08150 2 03360 00083 06282 09409 2 03127 001271990 04569 08155 2 03586 00079 06369 09424 2 03055 001241991 04289 08031 2 03743 00078 06092 09418 2 03326 001241992 04330 07957 2 03627 00078 06273 09299 2 03027 001241993 04557 07918 2 03362 00078 06428 09356 2 02929 001221994 04796 07921 2 03125 00079 06934 09312 2 02378 001211995 05147 07971 2 02825 00081 07221 09340 2 02119 001231996 05396 07938 2 02543 00085 07476 09290 2 01813 00119

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS) and are weighted

Restrictions Both samples include 19ndash44 year-old single women (divorced widowed or never married)who are not in school The March CPS sample excludes disabled or ill women and those with positive earnedincome but zero hours of work In the second panel single mothers without a child under six are excludedSee text for details

1083THE LABOR SUPPLY OF SINGLE MOTHERS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

ing at all in a given tax year but for some could decrease weeksworked If our goal is to provide a sharp test of theoretical pre-dictions whether a woman worked last year is a better outcomemeasure We report both measures with the expectation that theeffects of many of the recent policy changes on weekly employ-ment will be smaller than on annual employment

The employment rates reported in Table II exhibit a strikingtime pattern For single mothers weekly employment increasedby almost 6 percentage points between 1984 and 1996 whileannual employment increased over 85 percentage points Most ofthis increase occurred between 1991 and 199618 Focusing on thesubsample of single mothers with young children the employ-ment increases were even larger 10 percentage points for weeklyemployment and 135 percentage points for annual employmentIn contrast the declines in both weekly and annual employmentof about one percentage point for single women without childrensuggest that the rising employment of single mothers was not aresult of better work opportunities for all single women More-over the timing of the employment increases suggest that policychanges in the 1990s are likely to have played a large role

B Comparing Single Mothers and Single Women withoutChildren

Appendix 2 reports descriptive statistics for single womenwith and without children for the years 1984 1988 1992 and1996 The table indicates that single mothers tend to be older andless educated and are more likely to be nonwhite than singlewomen without children The age of single women without chil-dren rises appreciably over the sample period as does the edu-cation level of single mothers The fraction of single mothersliving with parents is stable while the rate for single womenwithout children falls The rates of cohabitation rise for bothsingle women with and without children

A potential criticism of the Table II results (and our main

18 One concern in interpreting changes in employment for single mothersduring the years 1992 to 1994 is that beginning in January 1994 the CPS used aredesigned questionnaire For a description of this CPS redesign see CohanyPolivka and Rothgeb [1994] and Polivka and Miller [1998] In Meyer and Rosen-baum [1999] we assess the extent of any bias due to the redesign using the parallelsurvey which provides contemporaneous responses using the new and old surveysWe also employ ORGMarch comparisons using the fact that redesign affected thetwo data sets at a different point in time Overall these comparisons indicate thatthe CPS redesign had a small effect that if it leads to any bias suggests that weslightly understate the recent employment increases of single mothers

1084 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

regression results below) is that single women without childrenare not a good comparison group for single mothers The means inAppendix 2 suggest the possibility that changes in the character-istics of single mothers versus single women without childrencould explain the two groupsrsquo differing employment rate trendsIn our regression results we condition on observable characteris-tics such as race and education in order to make the two groupsmore comparable It is also interesting to note that single womenwith and without children are quite similar in an importantdimension hourly earnings The mean hourly earnings of womenwith and without children are fairly similar (and they are muchcloser if one controls for education)

Perhaps more importantly one might argue that employ-ment rates are so high for single women without children that itis unreasonable to expect this group to respond to changes ineconomic conditions in the same way that single mothers do Yetemployment rates are not particularly high for low-educated sin-gle women particularly when examining employment last weekOnly 33 percent of high school dropout single mothers workedand 48 percent of high school dropout single women withoutchildren worked last week Nevertheless in our later regressionsderivative estimates for our key policy variables tend to be thelargest and most statistically signicant for high school dropouts

One might also wonder whether the large increases in em-ployment that we nd for single mothers but not for singlewomen without children also occur for other demographicgroups In Meyer and Rosenbaum [2000a] we examine whetherthere are similar employment increases for two other groups withhistorically low employment rates black males 19ndash44 and mar-ried mothers 19ndash44 We nd that the large increases in employ-ment of single mothers over 1984 ndash1996 and particularly since1991ndash1996 are not mirrored by other demographic groups

Another potential criticism of our approach is that usingvariation across women in their marital status number of chil-dren and state of residence implicitly assumes that marriagefertility and migration decisions are exogenous to the policychanges that we examine The evidence on the effects of policychanges on these decisions is mixed making the exogeneity as-sumption more plausible For example in her recent reviewHoynes [1997] concludes ldquoTogether this evidence suggests thatmarriage decisions are not sensitive to nancial incentivesrdquo Shealso argues that ldquoOverall [the effects of welfare on out-of-wedlock

1085THE LABOR SUPPLY OF SINGLE MOTHERS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

births] are often insignicant and when they are not they aresmall [pp 129ndash130]rdquo On the other hand another recent reviewMoftt [1997] suggests that the weight of the evidence impliessome effect of welfare benets on marriage and fertility As tolocation Meyer [1999] concludes that there is a signicant butsmall effect of welfare on migration Overall it is likely thatendogenous single motherhood and location exert a small bias onour results

C Accounting for Individual and State Characteristics

As mentioned above the results in Table II could be partlyexplained by differential changes over time in characteristicssuch as age and education for single women with and withoutchildren Moreover business cycles may differentially affect sin-gle women with and without children thereby leading to employ-ment shifts unrelated to policy changes Consequently Table IIIpresents probit employment estimates for single women control-ling for demographic and business cycle changes We include alarge number of controls for differences between the two groupsand we include the unemployment rate as well as its interactionwith whether or not a woman has children The specication thatwe estimate is

(6) Pr(Eit 5 1) 5 F a Xit 1 b tYEARt

1 g t(YEARt p ANYCHILDRENi)

where Eit equals one if woman i from year t reports positive hoursworked in the reference week for the ORG (or the previous yearfor the March CPS) Xit is a vector that includes demographic andbusiness cycle variables YEARt is an indicator variable for yeart and ANYCHILDRENi equals one for a woman with childrenThe year dummies control for labor market trends in overallfemale employment and the X vector controls for demographicand business cycle effect differences between the groups espe-cially compositional shifts over time Thus differences between g t

coefcients give difference-in-differences estimates controlling forthese other factors These differences can be interpreted as esti-mates of the combined effect of changes in all factors affecting theemployment of single mothers relative to single women withoutchildren

The demographic and business cycle variables accounted forin Table III include controls for state race ethnicity age educa-

1086 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TABLE IIIPROBIT EMPLOYMENT PROBABILITY ESTIMATES FOR SINGLE WOMEN 1984ndash1996

Explanatory variable

ORG worked lastweek = 1

March CPS workedlast year = 1

(1) (2)

Averagederivative

Standarderror

Averagederivative

Standarderror

Any children p 1984 2 00797 00107 2 01087 00160Any children p 1985 2 00856 00105 2 01199 00156Any children p 1986 2 00857 00103 2 01144 00153Any children p 1987 2 00880 00099 2 01056 00144Any children p 1988 2 00837 00096 2 00918 00140Any children p 1989 2 00663 00094 2 00745 00131Any children p 1990 2 00788 00095 2 00832 00136Any children p 1991 2 00823 00102 2 00916 00151Any children p 1992 2 00747 00106 2 00706 00159Any children p 1993 2 00601 00101 2 00830 00153Any children p 1994 2 00538 00098 2 00388 00145Any children p 1995 2 00405 00096 2 00154 00143Any children p 1996 2 00121 00097 00042 00140Nonwhite 2 00902 00019 2 00727 00033Hispanic 2 00405 00030 2 00608 00033Age 19ndash24 2 00210 00024 2 00077 00055Age 25ndash29 00070 00024 2 00107 00095Age 35ndash39 2 00049 00026 00008 00052Age 40ndash44 2 00108 00028 00107 00116High school dropout 2 02161 00022 2 01512 00032Some college 00870 00019 00989 00055Bachelors 01441 00025 01755 00055Masters 01295 00040 01927 00095Divorced 2 00068 00028 00062 00052Widowed 2 01201 00080 2 01218 00116Any children p divorced 01154 00038 00720 00063Any children p widowed 00978 00097 01148 00137[ of children under 18 2 00404 00014 2 00325 00020[ of children under 6 2 00955 00020 2 00699 00027Pregnant z z 2 01333 00063Unearned income ($1000s) z z 2 00035 00003Central city z z 2 00230 00030State unemployment rate () 2 00113 00008 2 00101 00015Any children p state unemployment

rate () 00017 00010 00032 00017Number of observations 373662 119019

Sources The data are from the 1984ndash1996 Current Population Survey Outgoing Rotation Group File(ORG) and the 1985ndash1997 March Current Population Survey (March CPS)

Restrictions See Table II for sample restrictionsControls Additional controls include indicators for state year calendar month and calendar month

interacted with any children (ORG)Notes Unearned income includes interest dividend Social Security veteransrsquo benets and retirement

income The omitted group is white non-Hispanic age 30ndash34 never married and not pregnant (March CPS)She does not live in a central city (March CPS) and has only a high school education See text for details

1087THE LABOR SUPPLY OF SINGLE MOTHERS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

tion marital status marital status interacted with a childrenindicator the number of children under six and eighteen thestate unemployment rate the state unemployment rate inter-acted with a children indicator (for the March CPS only) controlsfor pregnancy central city and unearned income and (for theORG only) controls for month and month interacted with a chil-dren indicator Note that the difference-in-differences calculatedby subtracting one YEAR p ANYCHILDREN coefcient fromanother are hardly affected by including the controls19 For ex-ample between 1984 and 1996 the weekly employment of singlemothers relative to single women without children rises 71 per-centage points without controls and 68 percentage points withcontrols20 For annual employment the difference-in-differencesestimator for 1984 to 1996 suggests an 117 percentage pointincrease in the relative annual employment of single motherswithout controls and an 113 percentage point increase withcontrols Again most of the increase occurs between 1991 and1996 Therefore these difference-in-difference estimates suggesta potential role for policy changes especially since 1991

VI POLICY VARIABLES AND EMPLOYMENT USING OUR SIMPLE

STRUCTURAL MODEL

We now move on to our main approach that uses our simplestructural model to distinguish between the different policies andto provide estimates that have a clearer interpretation Whilesome of the estimates rely on comparisons of single mothers andsingle women with children over time other estimates use avariety of other sources of identifying variation in our key explana-tory variables In some specications the identifying variationcomes from differences in taxes and benets for families of dif-ferent sizes and in different states as well as changes in thesetaxes and benets over time and differences in state living costs

Table IV reports estimates of our structural model of the

19 Due to the difculty in gauging the magnitude of probit coefcient esti-mates instead we report derivatives of the probability of working with respect toeach of the explanatory variables averaged over the single mothers in the sampleThus differences in the average derivatives for the YEAR p ANYCHILDRENvariables give changes over time in the difference in employment between singlewomen with and without children analogous to the changes that can be calculatedfrom Table II

20 The ldquowithout controlsrdquo results come from a weighted probit including onlythe year dummies and YEAR p ANYCHILDREN interactions

1088 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

effects of tax and welfare policy on the probability that a womanworks These specications provide estimates of the parametersin expression (5) of Section II and can be used to obtain estimatesof the effects of the different policy changes during the 1984 ndash1996period These specications also provide coefcients that can beused to summarize the effects of a wide range of policies and thatcan be used to simulate other policies In addition to the variablesshown in Table IV each of these probits include the controlvariables reported in Table III (except for the YEAR p ANY-CHILDREN interactions) along with a large number of familycomposition variables listed in the table notes These controlvariables imply that we are not using simple differences acrossfamily types to identify our coefcients We are using changesover time or differences across states in how different families aretreated We focus rst on the full sample specications in columns(1) and (5)

All of the coefcients on the income variables have the signsthat are implied by our simple structural model and are signi-cantly different from zero21 Lower taxes and maximum welfarebenets increase employment while higher welfare benets if awoman works (due to lower implicit taxes on earnings) increaseemployment Rather than restricting the income variables toenter the worknonwork decision as a single expected incomevariable we have allowed the coefcients on the different compo-nents of income to differ It is thus encouraging that the coef-cients on the income tax and welfare variables have roughly thesame magnitude as expected The one exception to this rule isthat the coefcient on Welfare Benets if Work in the weeklyemployment equation is substantially larger than the other in-come coefcients

A Taxes

The Income Taxes if Work coefcient implies that a onethousand dollar reduction in income taxes if a woman worksincreases employment last week by 27 percentage points andincreases employment last year by 45 percentage points Both ofthese effects are strongly signicant These coefcients indicateelasticities of the participation rate with respect to the return to

21 We examined the importance of allowing for correlation among the errorterms at the level of state p year p ANYCHILDREN using STATA These standarderrors are very close to those without this correction for clustering

1089THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

PR

OB

ITE

ST

IMA

TE

SO

FT

HE

EF

FE

CT

OF

PO

LIC

YV

AR

IAB

LE

SO

NT

HE

EM

PL

OY

ME

NT

OF

SIN

GL

EW

OM

EN

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lan

ator

yva

riab

le

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

ucat

ion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

732

004

172

003

342

001

822

004

492

008

792

005

262

002

07in

$100

0sy

ear

(00

034)

(00

080)

(00

057)

(00

051)

(00

053)

(00

150)

(00

090)

(00

063)

Wel

fare

max

imu

mbe

net

20

0340

20

0425

20

0357

20

0218

20

0295

20

0579

20

0349

20

0068

in$1

000s

yea

r(0

002

4)(0

005

6)(0

004

0)(0

003

6)(0

003

8)(0

010

6)(0

006

3)(0

004

5)W

elfa

rebe

ne

tif

wor

k0

0772

006

540

0916

005

390

0571

011

770

0626

002

33in

$100

0sy

ear

(00

073)

(00

171)

(00

121)

(00

109)

(00

107)

(00

299)

(00

177)

(00

125)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

019

852

029

262

025

472

010

872

017

422

026

942

020

192

011

94if

wor

k(0

023

9)(0

052

2)(0

039

8)(0

036

3)(0

034

8)(0

094

4)(0

058

6)(0

041

7)M

edic

aid

ifw

ork

20

0096

20

0040

000

132

001

672

000

452

001

190

0007

20

0072

in$1

000s

yea

r(0

003

3)(0

006

6)(0

005

6)(0

005

5)(0

004

4)(0

010

9)(0

007

6)(0

005

9)W

aive

rmdashan

yti

me

lim

it0

0136

004

080

0192

20

0065

001

910

0256

000

720

0169

(Ind

icat

orva

riab

le)

(00

071)

(00

160)

(00

119)

(00

102)

(00

125)

(00

329)

(00

209)

(00

149)

Wai

vermdash

any

term

inat

ions

002

220

0355

003

540

0124

004

820

1174

005

320

0148

(Ind

icat

orva

riab

le)

(00

110)

(00

260)

(00

181)

(00

158)

(00

223)

(00

607)

(00

375)

(00

250)

Tra

inin

gmdashed

uca

tion

20

0805

20

0824

20

0715

20

0563

20

0759

20

0708

20

1222

20

0257

in$1

000s

yea

r(0

019

0)(0

043

0)(0

031

1)(0

028

3)(0

031

0)(0

084

9)(0

050

7)(0

037

6)

1090 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EIV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

wor

ked

last

year

=1

Yea

rsof

edu

cati

onY

ears

ofed

uca

tion

All

lt12

12gt

12A

lllt

1212

gt12

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashjo

bse

arch

oth

er0

0446

004

720

0607

002

600

0526

006

690

0643

002

84in

$100

0sy

ear

(00

117)

(00

272)

(00

190)

(00

175)

(00

192)

(00

528)

(00

317)

(00

236)

Chi

ldca

re0

0227

002

720

0190

002

260

0229

004

380

0175

001

94in

$100

0sy

ear

(00

065)

(00

148)

(00

104)

(00

104)

(00

100)

(00

287)

(00

164)

(00

119)

Sta

teu

nem

ploy

men

tra

te2

001

002

000

982

001

012

001

052

000

832

001

332

000

902

000

53in

perc

enta

gepo

ints

(00

007)

(00

020)

(00

013)

(00

010)

(00

014)

(00

044)

(00

024)

(00

016)

Any

chil

dren

pst

ate

unem

pra

te2

000

010

0009

20

0010

000

080

0011

000

422

000

090

0022

inpe

rcen

tage

poin

ts(0

000

9)(0

002

1)(0

001

4)(0

001

3)(0

001

4)(0

004

2)(0

002

4)(0

001

7)N

um

ber

ofob

serv

atio

ns37

366

251

146

134

432

188

084

119

019

159

9441

060

619

65

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spec

ica

tion

s(2

)an

d(6

)ar

ere

stri

cted

tohi

ghsc

hoo

ldr

opou

ts(

3)an

d(7

)to

high

sch

ool

grad

uate

san

d(4

)an

d(8

)to

thos

ew

ith

aned

ucat

ion

beyo

nd

hig

hsc

hool

C

ontr

ols

Inad

diti

onto

the

vari

able

sin

Tab

leII

I(e

xcep

tfo

rth

ein

tera

ctio

nsbe

twee

nye

aran

dan

ych

ildr

en)

the

foll

owin

gco

ntro

lsar

ein

clud

ed

indi

cato

rsfo

rst

ate

year

ca

len

dar

mon

tha

nd

cale

ndar

mon

thin

tera

cted

wit

han

ych

ildr

en(O

RG

)w

heth

erat

leas

ton

etw

oth

ree

and

four

orm

ore

child

ren

are

pote

ntia

llyA

FD

Cel

igib

lew

heth

erat

leas

ton

ean

dat

leas

ttw

och

ildr

enar

eE

ITC

elig

ible

and

wh

ethe

rat

leas

ton

ech

ild

isu

nder

six

unde

rth

ree

unde

rtw

oan

dun

der

one

Las

tco

ntin

uou

sva

riab

les

for

the

num

ber

ofch

ildr

enun

der

each

age

betw

een

one

and

nin

etee

nar

ein

clud

ed

Not

esI

nal

lspe

cic

atio

ns

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

one

stim

ated

sepa

rate

lyfo

rsi

ngl

ew

omen

wit

han

dw

itho

utch

ildr

enT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ngdi

ffer

ence

san

dal

ldo

llar

amou

nts

are

expr

esse

din

1996

doll

ars

See

App

endi

x1

for

spec

ic

indi

ces

used

and

othe

rde

tail

s

1091THE LABOR SUPPLY OF SINGLE MOTHERS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

work of 107 for any employment during the year and 083 forwork in an average week We also estimated specications withseparate coefcients on state and federal income taxes althoughfor brevity these full estimates are not reported here The resultsfor federal taxes were similar to all taxes while the derivative(standard error) for state income taxes was a large and signicant2 00336 (00083) in the ORG sample and a smaller and insigni-cant 2 00165 (00139) in the March sample Thus while the statetax estimates are much less precise and differ in the two samplesthey give the same message as the other tax coefcients ie thatthe labor supply of single mothers responds to taxes

B Welfare

The full sample specications of columns (1) and (5) alsoindicate substantial effects of welfare on employment A onethousand dollar reduction in the annual Welfare Maximum Bene-t (the AFDC plus Food Stamp benet a women receives if shedoes not work) increases employment last week by 34 percentagepoints and increases employment last year by 30 percentagepoints This calculation holds constant the other welfare vari-ables Welfare Benets if Work and Probability of AFDC Receiptif Work that generally change with the maximum benet TheWelfare Benets if Work effect is sizable implying that a onethousand dollar increase in benets when one works increasesemployment last week by 72 percentage points and last year by57 percentage points These estimates suggest substantial posi-tive employment effects of reductions in implicit tax rates andincreases in earnings disregards

The transaction costs or stigma of welfare receipt as mea-sured by the Probability of AFDC Receipt if Work variable isnegative and signicantly different from zero as expected (seeequation (5)) The magnitude of this coefcient can be gauged bycomparing it with the coefcients on the variables denominatedin thousands of dollars Such comparisons suggest a transactioncost of several thousand dollars with the exact number depend-ing on the employment measure and the income variable usedFor example using the Welfare Benets if Work coefcient in theORG sample yields a transaction cost estimate of $2571 whilethe March sample implies an estimate of $3051 This resultagrees with past studies as well as ethnographies that havetended to nd substantial transaction costs or stigma of welfarereceipt

1092 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

To assess the effect of cutting the AFDC benet one needs toincorporate the effects of all three of the welfare variables and theMedicaid if Work variable When the AFDC maximum benetand payment standard are cut they not only reduce benets ifone does not work but also reduce benets if one does work Theyalso decrease the likelihood that a working mother will be onwelfare at all thereby reducing both her Medicaid eligibility andher AFDC transaction and stigma costs When we do the fullcalculations we nd that a 10 percent cut in the maximumbenet ($324 annually) increases both the annual and weeklyemployment rate by about 10 percentage points

Despite a more detailed calculation of welfare incentivesthan most past work and the use of panel data techniques wethink there are important potential sources of bias in these esti-mates We should also note that by dividing the effect of welfareinto income when working and when not and by estimating aseparate term for transaction costsstigma we are putting thetheoretical predictions to a more severe test than most work Asdiscussed in Section IV the Welfare Benets if Work variable andthe Probability of AFDC Receipt if Work variable are more dif-cult to calculate precisely than our other variables The largercoefcient on the Welfare Benets if Work variable could also bedue to the scale of this variable being inappropriately low Theearnings distribution used to calculate expected benets putsmost of the weight on earnings levels where welfare benetswould be low or zero It is very likely that we should use anearnings distribution that puts greater weight in the left tailsince women who work while on welfare rarely report all of theirearnings to the welfare ofce [Edin and Lein 1997] The reasonsfor possible bias in the Probability of AFDC Receipt if Workvariable are similar The coefcients on these two variables tendto both be large in the same specications with their oppositesigns canceling each other out

C Medicaid

We nd little effect of Medicaid on the employment decisionsof single mothers Theory predicts that the Medicaid if Workvariable will have a positive effect on employment The variablehas the opposite effect from this prediction in both samplesalthough the coefcient estimates are small and usually are notsignicantly different from zero This result is not completelyunexpected given the weak and conicting ndings in past work

1093THE LABOR SUPPLY OF SINGLE MOTHERS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

Part of the difculty is the uncertainty about individual knowledgeof Medicaid rules and their valuation of the benets We havetried a large number of alternative specications none of whichindicates a large effect of Medicaid A full accounting of theseresults can be found in Meyer and Rosenbaum [2000b]

D Welfare Waivers and Time Limits

The AFDC waiver variables have the expected effect on em-ployment and their coefcients are signicantly different fromzero Both the implementation of a time limit on welfare receiptand the actual termination of benets under a work requirementor time limit waiver are predicted to increase employment bybetween 14 and 48 percentage points However until the lastyears of our sample the overall importance of such waivers issmall Even by 1994 only 5 percent of single mothers lived instates with a time limit and less than half of 1 percent lived instates that had begun to terminate benets

One should be cautious in interpreting the waiver coef-cients especially in attributing effects to the implementation ofparticular provisions of recent waivers or the termination of casesper se The perception of welfare changes by potential welfarerecipients the attitudes of case workers and differences in stateimplementation of policies likely play a large role in inuencingthe welfare caseload and consequently employment It is alsoeconometrically difcult to disentangle which provisions of awaiver are the most important since states typically imple-mented several changes to their AFDC programs under waiversat the same time The reported coefcients are partly the effect ofthe particular actions coded and partly a proxy for other changesgoing on in the states

Recognizing these limitations the strength of the evidencehere for a causal interpretation of the waiver results is muchgreater than in the studies of welfare caseloads First we useimplementation dates rather than application or approval dateswhich are at best loosely related to when provisions are enforcedSecond when we account for state intentions to reform welfare asindicated by whether or not a state has made a major waiverapplication this variable has little effect Third one or two yearleads of our time limit and termination variables have small andinsignicant coefcients suggesting that the provisions per serather than publicity or administrator attitudes lead to the em-ployment increases This result contrasts with those of Blank

1094 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

[1997] and Levine and Whitmore [1998] who found strong effectsof leads of waiver variables on caseloads

E Training and Child Care

The last three coefcient estimates in Table IV measure theemployment effects of expenditures on training and child careHigher expenditures on job search and other training and onchild care are associated with a higher employment rate for singlemothers Training expenditures on education have a negativeeffect that is signicant in both samples The job search coef-cients imply that an increase in expenditures of one thousanddollars (about two-thirds of average expenditures) would increasethe employment rate for single mothers without young childrenby over four percentage points Since single mothers withoutchildren young enough to exempt them from training programsmake up about half of all single mothers the overall effect wouldbe over two percentage points An increase in federal and statechild care expenditures of ve hundred dollars per single motherwith a child under six (slightly less than the mean in 1996) isassociated with about a one percentage point increase in bothweekly and annual employment These effects are quite substan-tial per dollar expended The training result on education is notsurprising given the weaker results in the literature on classroomtraining and the possible short-term effect on employment aswomen are in classrooms rather than jobs

F Results by Education Group

Table IV also reports separate estimates for the effects of thepolicy variables for three education groups less than high schoolhigh school and some college We would expect a priori that thepolicy variables which mostly capture taxes and benets receivedby low-income women would have the greatest effect on highschool dropouts less of an effect on those with a high schooldegree and even less of an effect on those with some college22

Overall the results by level of education are consistent with thehypothesized larger effects on the less educated The derivativestend to be much larger in absolute value for high school dropoutsthan they are in the full sample and much smaller for those with

22 The estimates use a xed wagehours distribution (that does not vary byeducation) to calculate the income and benet variables so that the explanatoryvariables are comparable across the columns

1095THE LABOR SUPPLY OF SINGLE MOTHERS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

some college than in the full sample For example a one thousanddollar cut in taxes (or increase in tax credits) for high schooldropout single women is predicted to increase their employmentby 42 percentage points in a typical week and increase work atall during the year by 88 percentage points The correspondingnumbers for those with some college education are 18 percentagepoints and 21 percentage points Many of the other policy vari-able derivatives also fall with increased education23

G Unemployment and Macroeconomic Conditions

Table IV also reports the coefcients on the state unemploy-ment rate and its interaction with a dummy variable for a singlewoman having children The unemployment rate is strongly sig-nicant and implies that for single women without children a onepercentage point increase in the unemployment rate is associatedwith a 10 percentage point decrease in employment in a typicalweek and a 08 percentage point decrease in work anytime duringthe year On the other hand the interaction of the unemploymentrate with being a single mother is small and not signicantlydifferent from zero The point estimates imply that a one percent-age point increase in the unemployment rate is associated withonly a 001 percentage point decrease in a typical week and 01percentage point increase any time during the year in theemployment of single mothers relative to single women withoutchildren These coefcients indicate a strong and similarresponsiveness of both groups of single women to the state of themacroeconomy This result is favorable for the use of singlewomen without children as a comparison group for single mothers

H Alternative Specications

Since many of the changes in policy notably welfare reformtook place in recent years and a well-publicized decline in thewelfare rolls began in 1994 we reestimate the full sample speci-cations of Table IV dropping the years 1994 ndash1996 along withthe waiver variables (which are nearly always zero through1993) The estimates from this shorter sample which are re-ported in columns (1) and (5) of Table V are very close to those

23 The derivatives might be lower for groups with higher levels of educationbecause their employment rates are higher leaving less room for increases inemployment However the drop in the magnitude of the policy variable deriva-tives with more education is greater than it is for other control variables such asthe unemployment rate

1096 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

PR

OB

ITE

STIM

AT

ES

OF

TH

EE

FF

EC

TO

FP

OL

ICY

VA

RIA

BL

ES

ON

TH

EE

MP

LO

YM

EN

TO

FS

ING

LE

WO

ME

NA

LT

ER

NA

TIV

ES

PE

CIF

ICA

TIO

NS

AV

ER

AG

ED

ER

IVA

TIV

E(S

TA

ND

AR

DE

RR

OR)

Exp

lana

tory

vari

able

OR

G

wor

ked

last

wee

k=

1M

arch

CP

Sw

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Inco

me

taxe

sif

wor

k2

002

892

001

352

005

012

000

552

004

302

002

532

005

582

003

05in

$100

0sy

ear

(00

045)

(00

087)

(00

051)

(00

084)

(00

074)

(00

129)

(00

087)

(00

133)

Lag

ged

inco

me

taxe

sif

wor

kz

zz

20

0263

zz

z2

001

75in

$100

0sy

ear

zz

z(0

009

2)z

zz

(00

148)

Wel

fare

max

imu

mbe

net

20

0375

20

0107

20

0464

20

0340

20

0291

20

0011

20

0445

20

0296

in$1

000s

yea

r(0

002

9)(0

006

5)(0

003

3)(0

002

4)(0

004

7)(0

009

2)(0

005

6)(0

003

8)W

elfa

rebe

ne

tif

wor

k0

0911

001

070

1027

007

660

0644

20

0080

20

0882

005

67in

$100

0sy

ear

(00

089)

(00

106)

(00

099)

(00

073)

(00

132)

(00

144)

(00

157)

(00

107)

Pro

babi

lity

ofA

FD

Cre

ceip

t2

021

042

010

442

024

102

019

502

020

562

008

372

026

652

017

17if

wor

k(0

029

7)(0

033

6)(0

032

7)(0

023

9)(0

043

3)(0

050

1)(0

051

8)(0

034

8)M

edic

aid

ifw

ork

20

0133

000

292

002

092

000

952

000

790

0055

20

0107

20

0045

in$1

000s

yea

r(0

004

6)(0

003

7)(0

004

5)(0

003

3)(0

005

9)(0

004

9)(0

006

6)(0

004

4)W

aive

rmdashan

yti

me

lim

itz

001

710

0204

001

28z

002

470

0350

001

84(I

ndic

ator

vari

able

)z

(00

080)

(00

104)

(00

071)

z(0

014

1)(0

019

7)(0

012

5)W

aive

rmdashan

yte

rmin

atio

nsz

001

600

0456

002

15z

003

550

0703

004

79(I

ndic

ator

vari

able

)z

(00

116)

(00

163)

(00

110)

z(0

023

8)(0

034

7)(0

022

3)

1097THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

(CO

NT

INU

ED

)

Exp

lana

tory

vari

able

OR

Gw

orke

dla

stw

eek

=1

Mar

chC

PS

w

orke

dla

stye

ar=

1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Tra

inin

gmdashed

uca

tion

20

0755

20

0764

20

0728

20

0841

20

0804

20

0656

20

1396

20

0780

in$1

000s

yea

r(0

023

7)(0

022

3)(0

039

2)(0

019

0)(0

040

1)(0

036

5)(0

060

7)(0

031

0)T

rain

ingmdash

job

sear

cho

ther

007

620

0441

001

060

0428

007

150

0571

008

490

0517

in$1

000s

yea

r(0

014

0)(0

013

3)(0

023

7)(0

011

7)(0

023

7)(0

021

3)(0

038

2)(0

019

3)C

hild

care

000

110

0190

000

710

0215

20

0033

001

710

0286

002

22in

$100

0sy

ear

(00

092)

(00

073)

(00

088)

(00

066)

(00

140)

(00

114)

(00

161)

(00

101)

1994

ndash199

6ex

clud

edY

esN

oN

oN

oY

esN

oN

oN

oS

ingl

em

oth

ers

only

No

Yes

No

No

No

Yes

No

No

Mot

her

sw

och

ildr

enlt

6ex

cl

No

No

Yes

No

No

No

Yes

No

Nu

mbe

rof

obse

rvat

ions

292

731

122

966

303

396

373

662

938

1640

818

957

8611

901

9

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Res

tric

tion

sS

eeT

able

IIfo

rsa

mpl

ere

stri

ctio

ns

Spe

cic

atio

ns(1

)an

d(5

)ex

clu

de19

94ndash1

996

Spe

cic

atio

ns

(2)

and

(6)

incl

ude

only

sin

gle

mot

hers

Spe

cic

atio

ns

(3)

and

(7)

incl

ude

sing

lew

omen

wit

hout

chil

dren

and

sing

lem

othe

rsw

hose

youn

gest

chil

dis

less

than

six

Con

trol

sS

eeT

able

IVfo

rco

ntr

ols

Indi

cato

rsfo

rin

tera

ctio

nsbe

twee

nst

ate

and

any

AF

DC

elig

ible

chil

dren

and

betw

een

year

and

both

any

AF

DC

elig

ible

child

ren

and

any

EIT

Cel

igib

lech

ildre

nar

ein

clud

edin

spec

ica

tion

s(2

)an

d(6

)N

otes

In

alls

peci

cat

ions

the

tax

wel

fare

and

Med

icai

dva

riab

les

are

calc

ula

ted

usin

ga

join

tho

urs

wag

edi

stri

buti

ones

tim

ated

sepa

rate

lyfo

rsi

ngle

wom

enw

ith

and

wit

hout

chil

dren

Tax

esan

dw

elfa

rear

ead

just

edfo

rst

ate

cost

ofli

vin

gdi

ffer

ence

sA

lso

alld

olla

ram

ount

sar

eex

pres

sed

in19

96do

llar

sS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1098 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

over the full sample period The only exception to this generali-zation is that the job searchother training coefcient is largerand the child care coefcient is smaller and statistically insigni-cant over the shorter time span These results are among themost important in this paper because they indicate that (1) theurry of welfare reform measures after 1993 has not falsely led toour main results and (2) the extended recovery of the 1990s is notan alternative explanation for our main results

Next we examine a sample of only single mothers Thisspecication identies the effects of the income variables throughchanges across states and for different family sizes In the case ofthe Income Taxes if Work variable we are largely using thevariation from the last few years when the EITC for women withone child was nearly unchanged but the EITC for women with twoor more children rose in large steps Thus identication comesfrom using women with one child as a control group and chang-ing the treatment that women with two or more children receiveWith single mothers only the year indicators remove the timetrend in welfare receipt and benets and the state indicatorsremove time-constant differences in state welfare benets andmuch of these state cost of living differences in the income vari-ables Thus the variation in welfare benets used to identify thecoefcients is now changes in state-level benets This identi-cation approach examines the employment response to fairlysubtle or short-run features of the welfare and tax laws Thesepolicy changes may be overwhelmed by other factors in thesespecications Despite these potential difculties much of theincome tax effect remains although the estimates are muchsmaller While the effect of taxes is still signicant in the MarchCPS data the drop in the coefcient and larger standard errorleads the ORG coefcient to be insignicantly different from zeroThe welfare benet coefcients are now no longer signicant TheAFDC transaction cost coefcient however remains signicantin the ORG data while the Medicaid coefcient has the expectedsign but remains small and insignicant in both samples

In the third set of specications of Table V we only includesingle mothers with a child under six (and single women withoutchildren) The derivative estimates for the tax and welfare vari-ables including waivers are often substantially larger in magni-tude for these single mothers with young children especially forthe tax variable in the ORG sample These specications are ofparticular interest because the effects of increased employment

1099THE LABOR SUPPLY OF SINGLE MOTHERS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

on parental care is likely to be largest on these families withyoung children who are not likely to be in school The last speci-cation of Table V examines whether women learn about taxchanges with a delay after they are implemented This specica-tion includes both the contemporaneous and one year lagged taxvariables The results are somewhat supportive of a lagged effectof taxes In the ORG data the contemporaneous tax variable issmall and insignicant while the lagged variable is large andsignicant In the March CPS it is the contemporaneous variablethat is large and signicant while the lagged variable is smallerand signicant although still substantial in size

I Additional Specications and Hours Worked

We examine several other specications that are not reportedhere in order to determine the benets of studying many pro-grams at the same time to check the sensitivity of our results toalternative specications and to see whether there are particu-larly large effects for certain subgroups of the population We ndthat ignoring some of the policy changes that we study has asubstantial effect on the estimates for the remaining programsWhen we include the tax variable but leave out the other policyvariables its coefcient is about 50 percent larger in both sam-ples When the only policy variables that we include are Medicaidif Work and the Welfare Maximum Benet the Medicaid coef-cient is positive and signicant in the March CPS sample Whenthe other policy variables are not included the waiver variablesare much larger On the other hand the tax coefcient is hardlychanged when the training and child care variables are excludedThese results suggest that the common research strategy of in-vestigating one program in isolation has the potential to givemisleading results

We have examined the sensitivity of our results to alterna-tive samples and variable denitions In particular the resultsare little changed by using more stringent denitions of employ-ment by including separated women or women in school We alsotry several subgroup analyses In particular we examine differ-ences between whites and nonwhites and family heads and sub-family heads Nonwhites appear to be more affected by welfarewaivers than whites while subfamily heads are more sensitive totaxes than family heads

To obtain a broader picture of the effects of welfare and taxpolicy on labor supply we also examined hours worked (see

1100 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

Meyer and Rosenbaum [1999] for more details) Difference-in-differences estimates for hours analogous to those in Table IIshow large relative increases in work for single mothers over thesample period with almost all of the change occurring after 1991We also estimated a series of Tobit and OLS regressions todetermine the effects of tax and welfare policy on hours controllingfor demographics economic conditions state and year We in-clude the same variables as we did in Tables IV and V althoughwe should emphasize that these variables were constructed forour structural model of employment and so are less suitable foran analysis of hours The effects of the policy variables in theTobit estimates for all women whether or not they work tend to besimilar to the effects on employment seen in the earlier tablesThese results hold for the sample of single mothers as well as forall single women The results are very similar for hours per yearin the March CPS and hours in a typical week in the ORG Forhours worked conditioning on positive hours the policy variablestend to have much the same signs but smaller and less signi-cant coefcients Overall the results tend to conrm the resultsfor the main policy variables that we found in the employmentprobits

VII WHICH POLICIES ACCOUNTED FOR THE EMPLOYMENT CHANGES

Our simultaneous examination of many government policiesmakes it straightforward to estimate the relative contribution ofthese policies to the recent increase in employment of singlemothers In Table VI we decompose the employment increases forsingle mothers relative to single women without children for boththe entire period (1984 ndash1996) and the recent period of rapidemployment growth (1992ndash1996) Overall these decompositionsindicate a large role for the EITC and other tax changes modestroles for AFDC benet cuts and waivers and smaller roles forMedicaid training and child care increases

Using the parameter estimates from our main specications(specications (1) and (5) of Table IV) the EITC explains 62percent of the increase in weekly employment over the full 1984to 1996 period yet only 27 percent of the increase between 1992and 1996 For annual employment the EITC plays a very similarrole explaining 61 percent of the 1984 to 1996 increase and 35percent of the 1992 to 1996 increase The corresponding changesin employment attributed to the EITC over the full 1984 to 1996

1101THE LABOR SUPPLY OF SINGLE MOTHERS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

TA

BL

EV

IC

ON

TR

IBU

TIO

NO

FP

OL

ICY

CH

AN

GE

ST

OT

HE

CH

AN

GE

SIN

TH

ER

EL

AT

IVE

EM

PL

OY

ME

NT

OF

SIN

GL

EM

OT

HE

RS

VE

RSU

SS

ING

LE

WO

ME

NW

ITH

OU

TC

HIL

DR

EN

19

84ndash1

996

AN

D19

92ndash1

996

Exp

lan

ator

yva

riab

le

1984

ndash199

619

92ndash1

996

OR

GM

arch

CP

SO

RG

Mar

chC

PS

Din

emp

of

tota

lD

inem

p

ofto

tal

Din

emp

of

tota

lD

inem

p

ofto

tal

Inco

me

taxe

sif

wor

k0

0438

622

0

0720

614

0

0186

268

0

0305

351

W

elfa

rem

axim

um

ben

et

001

7925

4

001

5613

3

001

1416

5

000

9911

4

Wel

fare

ben

et

ifw

ork

000

050

70

0003

03

20

0045

26

52

000

332

38

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k2

000

022

03

20

0002

20

10

0020

29

000

172

0M

edic

aid

ifw

ork

20

0070

29

92

000

322

28

20

0023

23

32

000

112

12

Tot

alw

elfa

rebe

ne

tsamp

Med

icai

d0

0112

159

0

0125

106

0

0066

96

000

738

4W

aive

rmdashan

yti

me

lim

it0

0054

76

000

756

40

0052

76

000

738

5W

aive

rmdashan

yte

rmin

atio

ns

000

466

50

0099

84

000

466

60

0099

114

T

otal

wel

fare

wai

vers

000

9914

1

001

7414

8

000

9814

2

001

7219

8

Tra

inin

gmdashed

uca

tion

20

0101

214

4

20

0096

28

22

000

212

30

20

0020

22

3T

rain

ingmdash

job

sear

cho

ther

000

659

20

0077

65

000

476

80

0056

64

Ch

ild

care

000

689

70

0069

59

000

111

80

0013

15

Tot

altr

ain

ing

ampch

ild

care

000

324

60

0050

43

000

395

70

0039

56

Dem

ogra

phic

s2

000

732

104

2

000

682

58

001

0715

5

001

1212

9

Oth

er0

0096

137

0

0172

147

0

0196

283

0

0158

182

T

otal

007

0510

00

011

7410

00

006

9110

00

008

6910

00

Sou

rces

The

data

are

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

up

Fil

e(O

RG

)an

dth

e19

85ndash1

997

Mar

chC

urre

ntP

opul

atio

nS

urve

y(M

arch

CP

S)

Not

es

Din

emp

give

sth

ech

ange

inth

eem

ploy

men

tof

sing

lem

othe

rs(r

elat

ive

tosi

ngle

wom

enw

itho

ut

chil

dren

)ov

erth

esp

eci

edti

me

peri

odth

atis

due

toth

egi

ven

expl

anat

ory

vari

able

(s)

of

tota

lgiv

esth

epe

rcen

tage

ofth

eem

ploy

men

tin

crea

sefo

rsi

ngle

mot

hers

(rel

ativ

eto

sing

lew

omen

wit

hout

chil

dren

)exp

lain

edby

the

give

nex

plan

ator

yva

riab

le(s

)T

he

rela

tive

empl

oym

ent

incr

ease

sar

eth

eav

erag

ede

riva

tive

esti

mat

esof

the

inte

ract

ions

ina

wei

ghte

dpr

obit

mod

elin

clu

ding

year

indi

cato

rsan

dth

eir

inte

ract

ions

wit

han

any

chil

dren

indi

cato

rT

he

para

met

eres

tim

ates

used

toes

tim

ate

the

chan

gein

empl

oym

ent

com

efr

omsp

eci

cati

ons

(1)

and

(5)o

fTab

leIV

wh

ileth

ech

ange

over

tim

ein

the

polic

yva

riab

les

com

esfr

omA

ppen

dix

2

1102 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

period and the 1992 to 1996 period are also reported in Table VIWe estimate that the EITC and other tax changes increasedweekly employment 44 percentage points and annual employ-ment 72 percentage points over the full period with about 40percent of this change occurring over the 1992 to 1996 subperiodWhile these estimates are substantial they bracket the EITCeffects found by Eissa and Liebman [1996] and are smaller thanthose predicted by Dickert Houser and Scholz [1995] and Keane[1995]24

Changes in the maximum welfare benet and implicit taxrates and the Medicaid expansions account for between 10 and 16percent of the increase in weekly employment and between 8 and11 percent of the increase in annual employment over eitherperiod The effect of the Medicaid expansions themselves is usu-ally small or negative Conversely the effects of welfare waiversappear to be substantial with the estimates suggesting thatpolicies instituted under waivers account for about 14 to 15 per-cent of the increase in employment over the full sample periodand about 14 to 20 percent of the increase between 1992 and 1996for both weekly and annual employment In general both jobtraining and child care explain small parts of the employmentincrease although in the case of weekly employment over the fullperiod child care can account for about 10 percent of the increase

Improved macroeconomic conditions increased employmentfor both single mothers and single women without children overthe 1984 ndash1996 period Because the above calculations are forsingle mothers compared with single women without childrenunemployment is not given a share in the decomposition In all ofthe employment probits the interaction of unemployment andbeing a single mother had an economically small and statisticallyinsignicant effect Changes in state unemployment rates areestimated to have increased the absolute level of employment ofsingle mothers by 20 percentage points during a typical week

24 Eissa and Liebman [1996] found up to a 28 percentage point increase inparticipation due to TRA86 (which as we indicate in Section IV accounted for 43percent of the 1984ndash1996 change in taxes) Keane [1995] predicted that the1984ndash1996 changes will result in a 107 percentage point increase in participa-tion while Dickert Houser and Scholz [1995] predicted that the 1993ndash1996changes (39 percent of the full 1984ndash1996 change in taxes) would increase em-ployment of single parents by 33 percentage points Experimental ndings suchas those reported in Blank Card and Robins [2000] suggest substantial respon-siveness of welfare recipients and other low-income people to nancial incentivesThese experimental results would need to be extrapolated to all single mothersand the EITC to provide comparisons

1103THE LABOR SUPPLY OF SINGLE MOTHERS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

and 14 percentage points during the year over the 1984 ndash1996period These numbers are equivalent to 284 percent of therelative increase in weekly employment and 122 percent of therelative increase in annual employment of single mothers overthe period

In results not shown we recalculate the shares of the em-ployment increase due to various policies using the parameterestimates from specications with only single mothers (specica-tions (2) and (6) of Table V) These results suggest a much smallerrole for the EITC and other tax changes in explaining the changesin employment ranging from 49 to 56 percent as large as those inTable VI Changes in the maximum welfare benet are lessimportant while the results for welfare waivers job training andchild care are largely unchanged

VIII CONCLUSIONS

Between 1984 and 1996 tax and transfer policy were reori-ented to encourage work by single mothers Single mothers haveresponded to these incentives by working more especially after1991 and especially those with children under six To assesswhich policy changes have led to the employment increases weexamine the incentives of federal and state income taxes AFDCMedicaid Food Stamps and their implicit tax rates and earningsdisregards as well as AFDC waivers instituting time limits orwork requirements Our detailed examination of these policychanges using two large micro data sets indicates that EITC andother tax changes account for over 60 percent of the 1984 to 1996increase in the weekly and annual employment of single mothersrelative to single women without children Changes to welfareprograms were less important but still account for a substantialshare of the employment increases Changes in Medicaid train-ing and child care programs play a considerably smaller roleThese ndings are conrmed in an analysis of hours worked

This paper makes several methodological improvements overpast research including the estimation of a simple structuralmodel of employment which provides several independent tests ofthe hypothesis that single mothers respond to economic incen-tives Our results indicate that nancial incentives have powerfuleffects on single mothersrsquo employment decisions and that thedifferent sources of these incentives have effects of plausiblemagnitudes We also nd a sizable transaction cost or stigma to

1104 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

welfare We rely on less subjective measures of welfare waiverssuch as implementation dates and the beginning of case termi-nations and provide the rst evidence on the effects of waivers onemployment Unlike most past work we examine the major pro-grams affecting single mothers together nding that examiningone or two programs in isolation can lead to biases in estimatedbehavioral effects

In most of our specications identication comes from thedifferences in incentives faced by single women with and withoutchildren While we argue that single women without children area plausible comparison group we also provide estimates that donot rely on this comparison Instead these estimates rely onchanges in the treatment of family size state cost of living dif-ferences changes in state income taxes differences in earningsdisregards and implicit tax rates across states and changes inthese parameters and welfare benets within a state over timeOur nding of large tax and welfare effects on employment arerobust although tax effects and especially welfare effects aresometimes smaller using alternative identication strategies

Our result that the EITC played a dominant role in theemployment increases of single mothers between 1984 and 1996suggests that policies that ldquomake work payrdquo are effective in in-creasing work by single mothers This lesson is important in lightof the emphasis on punitive measures such as time limits andwork requirements in the most recent welfare reforms

APPENDIX 1 DESCRIPTION OF POLICY VARIABLES

This section describes the construction of our policy variablesand lists our information sources First we begin with the as-sumptions that we use to determine taxes program participationand benet levels

1 The determination of whether a woman has children andhow many she has is based on the CPS family and sub-family denitions Children in primary families (both re-lated and unrelated) are assigned to the family headwhile children in subfamilies are assigned to the subfam-ily head rather than to the primary family head Childrenare dened as any member of the given family (primary orsubfamily) under age 19 (or under 24 and a full-timestudent) for EITC purposes and under age 18 for all otherprograms

1105THE LABOR SUPPLY OF SINGLE MOTHERS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

2 In the March CPS sample the age for tax purposes is theage at the time of the March interview We subtract onefor AFDC and Medicaid purposes In the ORG sample weuse the age at the time of the interview for AFDC andMedicaid but for tax purposes we add one for interviewsoccurring between January and June

3 Women have no unearned income (including child sup-port) or assets and their children have no earned incomeunearned income or assets hence earnings determinetheir program eligibility

4 Single mothers are assumed to le as head of householdand claim their children as dependents while singlewomen without children le as single Also all womentake the standard deduction

5 Women receiving AFDC are in their rst four months ofwork and do not claim child care expenses25

6 Single women without children do not receive FoodStamps

7 Shelter costs (an input in Food Stamp calculations) varyonly by state and over time

A Tax Welfare and Medicaid Variables

First for each woman we calculate ve quantities income taxliabilities (federal and state income taxes incorporating federaland state EITCs) welfare benets (AFDC plus Food Stamps)AFDC receipt (indicator for AFDC eligibility) and Medicaidadults covered and Medicaid children covered Under the assump-tions above these calculations are made at 50 annual earningslevels generated from the cells of a joint wagehours distributionThe 50 cells come from a combination of ve annual hours levels(500 1000 1500 2000 and 2500) and ten hourly wage levels (45 6 7 8 10 12 15 20 and 25)

Second we use the wagehours distributions described in thetext to weight the above quantities We calculate the distributionsusing only women with more than $500 of annual earnings Wethen construct the following variables

25 These assumptions are roughly consistent with the facts In scal year1995 over two-thirds of AFDC families with earnings were in their rst fourmonths of work and only about 16 percent of AFDC families with earningsclaimed child care expenses [U S Department of Health and Human ServicesCharacteristics of AFDC Recipients 1996]

1106 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

c Income Taxes if Work is the weighted sum of income taxliabilities at the various annual earnings points using thewagehours distributions described above as weights

c Welfare Benets if Work is the weighted sum of welfarebenets at the various annual earnings points using thewagehours distributions described above as weights

c Probability of AFDC if Work is the weighted sum of AFDCreceipt at the various annual earnings points using thewagehours distributions described above as weights

c Medicaid if Work is calculated in two steps First wecalculate the weighted sum of Medicaid adults covered andMedicaid children covered at the various annual earningspoints using the wagehours distributions described aboveas weights Second we then multiply these sums by dollarexpenditures separately for adults and children In themain specications we use average expenditures over allstates and years

c Welfare Maximum Benet is the welfare benet assumingzero earnings

We calculate AFDC monthly benets (AFDC) as follows (settingquantities in parentheses to zero if negative)

(A1) AFDC 5 min MAXBEN RR p [PS 2 BRR p (EI 2 DIS)]

wheremdash MAXBEN is the maximum benetmdash RR is the ratable reductionmdash PS is the payment standard (the dollar amount when

benets end not counting disregards)mdash BRR is the benet reduction ratemdash EI is earned income andmdash DIS is the earnings disregard

We calculate Food Stamp benets in two steps (setting quantitiesin parentheses to zero if negative) First we calculate themonthly shelter cost expense deduction (SED) and second wecalculate the monthly Food Stamp benet (FS)

(A2) SED 5 (min SEDC SE 2 05

p ((1 2 EIDP) p EI 1 AFDC 2 SD))

(A3) FS 5 (MB 2 03 p ((1 2 EIDP)

p EI 1 AFDC 2 SD 2 SED))

1107THE LABOR SUPPLY OF SINGLE MOTHERS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

wheremdash EIDP is the earned income deduction percentage (018

prior to 1986 020 starting in 1986)mdash MB is the maximum Food Stamp benetmdash SD is the standard deductionmdash SE is shelter expensesmdash SEDC is the shelter expense deduction ceilingTax and welfare variables (and earned and unearned income

variables) are adjusted for state cost of living differences usingthe poverty threshold index for 1990 from National ResearchCouncil [1995] which is adjusted annually using the PCE dea-tor The poverty threshold index accounts for housing cost differ-ences between states using Census housing cost data

Sources for Taxes Welfare and Medicaid

We obtain the federal income tax schedules from the U SDepartment of the Treasury [various years] The state tax infor-mation was obtained from four sources the Advisory Committeeon Intergovernmental Relations [various years] the CommerceClearing House [various years] unpublished data from the Cen-ter on Budget Policy and Priorities and Feenberg and Coutts[1993] The AFDC program parameters are obtained from theU S Department of Health and Human Services (Characteristicsof State Plans [various years]) and unpublished data from theUrban Institute The Food Stamp parameters come from the U SHouse of Representatives (Green Book [various years]) and theU S Department of Agriculture [various years] The Medicaidprogram information is obtained from three sources the NationalGovernorrsquos Association [various dates] the IntergovernmentalHealth Policy Project [various years] and the U S House ofRepresentatives [Medicaid Source Book 1988 1993] Medicaiddollar values (separately for adults and children) come from un-published tables from the Health Care Financing Administration(HCFA)

B Welfare Waiver Variables

c Any Time Limit is one starting with the implementationmonth of a waiver that imposes mandatory work require-ments on families that reach time limits or results in thereduction or total loss of AFDC payments after a certaintime limit has been reached (usually two years)

1108 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

c Any Terminations is one beginning with the month inwhich a case is rst terminated under a welfare waiver

c Major Waiver Application is one beginning with the monthin which a state rst applies for a major statewide waiver

Note that these variables are always zero for women withoutAFDC children

Sources for Welfare Waiver Variables

The waiver variables we used are based on our reading of thewaiver summaries in General Accounting Ofce [1997] the U SDepartment of Health and Human Services [1997c] and Savnerand Greenberg [1997] These sources generally have the imple-mentation dates of waivers We also consulted American PublicWelfare Association [1996] Levine and Whitmore [1998] andU S Department of Health and Human Services [1997a] Ourclassication scheme follows most closely the classicationschemes in General Accounting Ofce [1997] and the U S De-partment of Health and Human Services [1997c]

C Training Program Variables

These variables measure variation across states and overtime in federal and state spending on welfare-to-work programsand on eligibility criteria These numbers are based on the statelevel scal year WIN (Work Incentive) program expenditures andstate level scal year JOBS (Job Opportunities and Basic Skills)program expenditures by component (job search education etc)We calculate spending per female AFDC adult who is not exemptfrom participation based on the age of her youngest child Thedollars are then divided by the state average wage to obtain anamount of services provided26

We calculate the distribution of the age of the youngest childand we apportion total JOBS spending to women using the frac-tion of participants who are female adults We divide spendinginto two categories education which includes education postsec-ondary education and self-initiated education and other whichincludes job search job development and placement on-the-jobtraining work supplementation community work experienceself-initiated training job skills job readiness and assessmentand employability plan For scal year 1990 it is necessary to

26 The state average wage is average hourly wage for manufacturing in thestate It is normalized so that the 1996 value = 100

1109THE LABOR SUPPLY OF SINGLE MOTHERS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

extrapolate WIN expenditures forward and JOBS expendituresbackward to the date when the JOBS program began in a givenstate We also extrapolate 1985 WIN data back to 1984 and scalyear 1996 forward to the rst three months of scal year 1997

Note that the training variables are zero for women withoutAFDC children and women with children young enough to ex-empt the mother from participation in WIN or JOBS

Sources for Training Variables

JOBSWIN expenditure data come from unpublished U SDepartment of Health and Human Services and U S Depart-ment of Labor tabulations and the U S House of Representa-tives (Green Book [various years]) To calculate the distribution ofthe age of youngest child for single mothers we use data from theU S Department of Health and Human Services (Characteristicsof AFDC Recipients [various years]) and authorsrsquo calculationsfrom the March CPS Wage data come from the Bureau of LaborStatistics web site

D Child Care Variable

Child Care expenditures are actual federal and state expen-ditures by state on the following four programs AFDC ChildCare Transitional Child Care At-Risk Child Care and ChildCare and Development Block Grants Expenditures are put on aper-person basis by dividing through by the number of unmarriedwomen with children less than six This denominator is calcu-lated using annual data on the number of women by state (fromthe Census Bureau) and the fraction of women in a state who areunmarried with children less than six which is calculated fromthe ORG over the entire 1984 ndash1996 period Like training dollarsthe resulting dollar value is then divided by the state averagewage to obtain an amount of services provided

Note that the child care variable is always zero for womenwithout children less than six

Sources for Child Care Variable

Child Care expenditures come from unpublished U S De-partment of Health and Human Service tabulations Annual dataon the number of women by state come from the U S CensusBureau The fraction of women in a state who are unmarried withchildren less than six is calculated from the ORG by the authorsWage data come from the Bureau of Labor Statistics web site

1110 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

AP

PE

ND

IX2

VA

RIA

BL

EM

EA

NS

FO

RS

ING

LE

MO

TH

ER

SA

ND

SIN

GL

EW

OM

EN

WIT

HO

UT

CH

ILD

RE

N1

984

1988

199

219

96

Var

iabl

e

1984

1988

1992

1996

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Ch

ildr

enN

och

ildr

enC

hil

dren

No

chil

dren

Age

314

426

86

319

727

79

319

628

83

321

929

22

Non

wh

ite

037

10

155

036

30

162

038

40

178

037

70

207

His

pan

ic0

086

005

30

103

007

20

111

007

90

136

009

3H

igh

sch

ool

drop

out

026

20

094

024

60

091

024

10

094

021

10

092

Som

eco

lleg

e0

211

029

70

234

030

50

256

031

70

311

031

7B

ach

elor

s0

063

019

20

064

020

40

061

021

00

072

023

3M

aste

rs0

022

005

90

025

006

10

023

006

60

021

006

4D

ivor

ced

056

40

151

053

30

161

047

70

165

046

00

162

Wid

owed

006

60

010

005

50

010

004

70

012

003

80

012

Liv

ing

wit

hpa

ren

ts0

156

041

80

151

037

50

154

034

70

154

033

9L

ivin

gw

ith

un

rela

ted

adu

ltm

ale

009

70

135

012

50

167

014

80

198

016

50

218

[of

chil

dren

un

der

181

681

z1

664

z1

707

z1

707

z[

ofch

ildr

enu

nde

r6

056

0z

057

1z

062

4z

061

3z

Ear

ned

inco

me

(Mar

ch)

128

1018

331

134

6220

183

132

3218

993

147

6019

912

Ear

ned

inco

me

ifw

ork

(Mar

ch)

172

5019

536

180

1321

293

179

3020

453

180

2321

479

An

nu

alho

urs

ifw

ork

(Mar

ch)

1718

1837

1724

1894

1712

1862

1740

1881

Hou

rly

earn

ings

(Mar

ch)

985

103

89

9110

86

987

105

910

03

109

8In

com

eta

xes

ifw

ork

1521

2751

1030

2958

811

2967

7929

14W

elfa

rem

axim

um

ben

et

7583

z74

06z

7391

z70

56z

Wel

fare

ben

et

ifw

ork

1482

z14

78z

1546

z14

88z

Pro

babi

lity

ofA

FD

Cre

ceip

tif

wor

k0

255

z0

287

z0

266

z0

256

zM

edic

aid

ifw

ork

1215

013

590

1704

219

424

Nu

mbe

rof

obse

rvat

ion

s93

9118

914

9211

186

1210

333

193

1187

8815

846

Sou

rces

The

data

are

prim

arily

from

the

1984

ndash199

6C

urre

ntP

opul

atio

nSu

rvey

Out

goin

gR

otat

ion

Gro

upF

ile(O

RG

)and

from

the

1985

ndash199

7M

arch

Cur

rent

Pop

ulat

ion

Surv

ey(M

arch

)R

estr

icti

ons

See

Tab

leII

for

sam

ple

rest

rict

ion

sN

otes

Mea

ns

com

efr

omth

eO

RG

unle

ssth

eyar

ela

bele

d(M

arch

)T

hese

mea

nsar

eca

lcul

ated

usi

ngth

ech

arac

teri

stic

sof

the

give

nsa

mpl

efo

rth

egi

ven

year

and

are

wei

ghte

dW

omen

are

assu

med

tobe

inth

eir

rst

four

mon

ths

ofw

ork

toha

ven

ou

near

ned

inco

me

and

tocl

aim

noch

ildca

reex

pens

esA

lso

sin

gle

wom

enw

ith

and

wit

hou

tch

ildr

enar

eas

sum

edto

le

ashe

adof

hous

ehol

dan

dsi

ngl

ere

spec

tive

ly

and

tocl

aim

the

stan

dard

dedu

ctio

nT

axes

and

wel

fare

are

adju

sted

for

stat

eco

stof

livi

ng

diff

eren

ces

All

doll

aram

ount

sar

eex

pres

sed

in19

96do

llar

spe

rye

arS

eeA

ppen

dix

1fo

rsp

eci

cin

dice

sus

edan

dot

her

deta

ils

1111THE LABOR SUPPLY OF SINGLE MOTHERS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

DEPARTMENT OF ECONOMICS AND INSTITUTE FOR POLICY RESEARCH NORTHWESTERN

UNIVERSITY AND NATIONAL BUREAU OF ECONOMIC RESEARCH

DEPARTMENT OF ECONOMICS UNIVERSITY OF NORTH CAROLINA GREENSBORO

REFERENCES

American Public Welfare Association ldquoSummary of Approved AFDC WaiverActionsrdquo (Washington DC May 1993 and August 1 1996)

Advisory Committee on Intergovernmental Relations Signicant Features ofFiscal Federalism (Washington DC various years)

Bavier Richard ldquoAn Early Look at the Effects of Welfare Reformrdquo unpublishedpaper March 1999

Blank Rebecca M ldquoWhat Causes Public Assistance Caseloads to Growrdquo unpub-lished paper May 1997

mdashmdash ldquoThe Effect of Medical Need and Medicaid on AFDC Participationrdquo Journalof Human Resources XXIV (1989) 54ndash87

Blank Rebecca M and Patricia Ruggles ldquoWhen Do Women Use Aid to Familieswith Dependent Children and Food Stampsrdquo Journal of Human ResourcesXXXI (1996) 57ndash89

Blank Rebecca M David E Card and Philip K Robins ldquoFinancial Incentives forIncreasing Work and Income among Low-Income Familiesrdquo in Finding JobsWork and Welfare Reform David Card and Rebecca M Blank eds (NewYork Russell Sage Foundation 2000) pp 373ndash419

Browning Edgar K ldquoEffects of the Earned Income Tax Credit on Income andWelfarerdquo National Tax Journal XLVIII (1995) 23ndash43

Cohany Sharon Anne Polivka and Jennifer Rothgeb ldquoRevisions in the CurrentPopulation Survey Effective January 1994rdquo Employment and Earnings XLI(1994) 13ndash37

Commerce Clearing House State Tax Handbook (Chicago IL various years)Council of Economic Advisers ldquoThe Economics of Child Carerdquo (Washington DC

December 1997)Danziger Sheldon Robert Haveman and Robert Plotnick ldquoHow Income Trans-

fers Affect Work Savings and the Income Distribution A Critical ReviewrdquoJournal of Economic Literature XIX (1981) 975ndash1028

Dickert Stacy Scott Houser and John Karl Scholz ldquoThe Earned Income TaxCredit and Transfer Programs A Study of Labor Market and Program Par-ticipationrdquo in Tax Policy and the Economy 9 James M Poterba ed (Cam-bridge MA MIT Press 1995) pp 1ndash50

Edin Kathryn and Laura Lein Making Ends Meet How Single Mothers SurviveWelfare and Low-Wage Work (New York Russell Sage Foundation 1997)

Eissa Nada and Jeffrey B Liebman ldquoLabor Supply Response to the EarnedIncome Tax Creditrdquo Quarterly Journal of Economics CXII (1996) 605ndash637

Eissa Nada and Hilary Williamson Hoynes ldquoThe Earned Income Tax Credit andLabor Supply Married Couplesrdquo NBER Working Paper No 6856 December1998

Ellwood David T ldquoThe Impact of the Earned Income Tax Credit and Social PolicyReforms on Work Marriage and Living Arrangementsrdquo National Tax JournalLIII (2000) 1063ndash1106

Feenberg Daniel and Elisabeth Coutts ldquoAn Introduction to the TAXSIM ModelrdquoJournal of Policy Analysis and Management XII (1993) 189ndash194

Fraker Thomas Robert Moftt and Douglas Wolf ldquoEffective Tax Rates andGuarantees in the AFDC Program 1967ndash1982rdquo Journal of Human Re-sources XX (1985) 252ndash263

General Accounting Ofce ldquoWelfare Reform Statesrsquo Early Experiences withBenet Terminationrdquo (Washington DC GAO S-97-74 May 1997)

Gueron Judith M and Edward Pauly From Welfare to Work (New York RussellSage Foundation 1991)

Heckman James J ldquoWhat Has Been Learned about Labor Supply in the PastTwenty Yearsrdquo American Economic Review LXXXIII (1993) 116ndash121

Hill Carolyn V Joseph Hotz Charles H Mullin and John Karl Scholz ldquoEITC

1112 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

Eligibility Participation and Compliance Rates for AFDC Households Evi-dence from the California Caseloadrdquo unpublished paper March 1999

Hoffman Saul D and Laurence S Seidman The Earned Income Tax CreditAntipoverty Effectiveness and Labor Market Effects (Kalamazoo MI UpjohnInstitute for Employment Research 1990)

Holtzblatt Janet Janet McCubbin and Robert Gillette ldquoPromoting Workthrough the EITCrdquo National Tax Journal XLVII (1994) 591ndash608

Hoynes Hilary Williamson ldquoWork and Marriage Incentives in Welfare ProgramsWhat Have We Learnedrdquo in Fiscal Policy Lessons from Economic ResearchAlan J Auerbach ed (Cambridge MA MIT Press 1997)

Intergovernmental Health Policy Project Major Changes in State Medicaid andIndigent Care Programs (Washington DC various years)

Jencks Christopher and Joseph Swingle ldquoHas Welfare Reform Helped or HurtSingle Momsrdquo unpublished paper February 2000

Keane Michael ldquoA New Idea for Welfare Reformrdquo Federal Reserve Bank ofMinneapolis Quarterly Review XXXVIII (1995) 2ndash28

Levine Judith A ldquoPolicy Incentives Confront Everyday Realities IntegratingEconomic and Sociological Perspectives on the Welfare-to-Work Transitionrdquounpublished paper March 1997

Levine Phillip B and Diane M Whitmore ldquoThe Impact of Welfare Reform on theAFDC Caseloadrdquo National Tax Association Proceedings Ninetieth AnnualConference (1998) 24ndash33

Liebman Jeffrey B ldquoThe Impact of the Earned Income Tax Credit on Incentivesand Income Distributionrdquo in Tax Policy and the Economy 12 James MPoterba ed (Cambridge MA MIT Press 1998) pp 83ndash120

Martini Alberto and Michael Wiseman ldquoExplaining the Recent Decline in Wel-fare Caseloads Is the Council of Economic Advisers Rightrdquo Challenge XL(1997) 6 ndash20

Meyer Bruce D ldquoDo the Poor Move to Receive Higher Welfare Benetsrdquo unpub-lished paper April 1999

Meyer Bruce D and Dan T Rosenbaum ldquoWelfare the Earned Income Tax Creditand the Labor Supply of Single Mothersrdquo NBER Working Paper No 7363September 1999

Meyer Bruce D and Dan T Rosenbaum ldquoMaking Single Mothers Work RecentChanges in Tax and Welfare Policy and their Effectsrdquo National Tax JournalLIII (2000a) 1027ndash1062

Meyer Bruce D and Dan T Rosenbaum ldquoMedicaid Private Health Insuranceand the Labor Supply of Single Mothersrdquo unpublished paper November2000b

Moftt Robert ldquoAn Economic Model of Welfare Stigmardquo American EconomicReview LXXIII (1983) 1023ndash1035

mdashmdash ldquoIncentive Effects of the U S Welfare Systemrdquo Journal of Economic Litera-ture XXX (1992) 1ndash61

mdashmdash ldquoThe Effect of Welfare on Marriage and Fertility What Do We Know andWhat Do We Need to Knowrdquo unpublished paper December 1997

Moftt Robert and Barbara Wolfe ldquoThe Effect of the Medicaid Program onWelfare Participation and Labor Supplyrdquo Review of Economics and StatisticsLXXIV (1992) 615ndash626

National Governorrsquos Association MCH (Maternal and Child Health) Update(Washington DC various dates)

National Research Council Measuring Poverty A New Approach Constance FCitro and Robert T Michael eds (Washington DC National Academy Press1995)

mdashmdash Evaluating Welfare Reform A Framework and Review of Current WorkRobert Moftt and Michele Ver Ploeg eds (Washington DC National Acad-emy Press 1999)

Polivka Anne E and Stephen M Miller ldquoThe CPS After the Redesign Refocus-ing the Economic Lensrdquo in Labor Statistics Measurement Issues John Halti-wanger Marilyn E Manser and Robert Topel eds (Chicago University ofChicago Press 1998) pp 249ndash286

Romich Jennifer L and Thomas Weisner ldquoHow Families View and Use the

1113THE LABOR SUPPLY OF SINGLE MOTHERS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS

EITC The Case for Lump-sum Deliveryrdquo National Tax Journal LIII (2000)1245ndash1265

Savner Steve and Mark Greenberg ldquoThe CLASP Guide to Welfare Waivers1997rdquo (Washington DC Center for Law and Social Policy 1997)

Scholz John Karl ldquoThe Participation Rate of the Earned Income Tax CreditrdquoInstitute for Research on Poverty Madison WI Discussion Paper 928-90October 1990

mdashmdash ldquoThe Earned Income Tax Credit Participation Compliance and AntipovertyEffectivenessrdquo National Tax Journal XLVII (1994) 59ndash81

Smeeding Timothy M Katherine E Ross Michael OrsquoConnor and Michael SimonldquoThe EITC Expectation Knowledge Use and Economic and Social MobilityrdquoNational Tax Journal LIII (2000) 1187ndash1210

U S Department of Agriculture Characteristics of Food Stamp Households(Alexandria VA U S Department of Agriculture Food and Consumer Ser-vice Ofce of Analysis and Evaluation various years)

U S Department of Health and Human Services Characteristics and FinancialCircumstances of AFDC Recipients (Washington DC U S Department ofHealth and Human Services Administration for Children and FamiliesOfce of Family Assistance Division of Performance Measurement variousyears)

mdashmdash Characteristics of State Plans for Aid to Families with Dependent Children(Washington DC U S Department of Health and Human Services Admin-istration for Children and Families Ofce of Family Assistance variousyears)

mdashmdash HHS Fact Sheet State Welfare Demonstrations (Washington DC U SDepartment of Health and Human Services March 17 1997a)

mdashmdash National Evaluation of Welfare-to-Work Strategies (Washington DC Ad-ministration for Children and Families Ofce of the Assistant Secretary forPlanning and Evaluation December 1997b)

mdashmdash Setting the Baseline A Report on State Welfare Waivers (Washington DCU S Department of Health and Human Services Ofce of the AssistantSecretary for Planning and Evaluation 1997c)

U S Department of the Treasury Statistics of Income Individual Income TaxReturns (Washington DC U S Department of the Treasury Internal Reve-nue Service various years)

U S Department of the Treasury Your Federal Income Tax Tax Guide forIndividuals (Washington DC U S Department of the Treasury InternalRevenue Service various years)

U S House of Representatives Committee on Energy and Commerce MedicaidSource Book Background Data and Analysis (Washington DC GovernmentPrinting Ofce November 1988 and January 1993)

U S House of Representatives Committee on Ways and Means Green BookBackground Material and Data on Programs within the Jurisdiction of theCommittee on Ways and Means (Washington DC Government Printing Of-ce various years)

Winkler Anne ldquoThe Incentive Effects of Medicaid on Womenrsquos Labor SupplyrdquoJournal of Human Resources XXVI (1991) 308ndash337

Yelowitz Aaron S ldquoThe Medicaid Notch Labor Supply and Welfare Participa-tionrdquo Quarterly Journal of Economics CX (1995) 909ndash940

Ziliak James P David N Figlio Elizabeth E Davis and Laura S ConnollyldquoAccounting for the Decline in AFDC Caseloads Welfare Reform or EconomicGrowthrdquo unpublished paper July 1997

1114 QUARTERLY JOURNAL OF ECONOMICS