WWW

2.09 V

V

94.35 tag

This is to certify that the

dissertation entitled

THE IMPACT OF THE EARNED INCOME TAX CREDIT ON
POVERTY, LABOR SUPPLY, AND HUMAN-CAPITAL ACCUMULATION

presented by

SCOTT B . DARRAGH

has been accepted towards fulﬁllment
of the requirements for

Ph . D . degree in Economics

ngﬂw

Major professor

 

 

 

MSU is an Afﬁrmative Action/Equal Opportunity Institution 0-12771

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuepBS-sz

THE IMPACT OF THE EARNED INCOME TAX CREDIT ON POVERTY, LABOR
SUPPLY, AND HUMAN-CAPITAL ACCUMULATION

By

Scott B. Darragh

A DISSERTATION
Submitted to
Michigan State University

in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Economics

2002

ABSTRACT

THE IMPACT OF THE EARNED INCOME TAX CREDIT ON POVERTY, LABOR
SUPPLY AND HUMAN-CAPITAL ACCUMULATION

By

Scott B. Darragh

The Earned Income Tax Credit (EITC) has become the largest public expenditure
program aimed at providing support to low-income, working families. This dissertation
evaluates how effective both the federal and state EITCs have been at increasing labor-
force participation and reducing poverty among adult women from 1982 to 1996.
Controls are included for other policies aimed at helping low-income families, such as
minimum wages and welfare waivers. Using pooled cross-sectional data from the March
CPS, the federal EITC is found to signiﬁcantly increase labor-force participation and
reduce poverty, especially among women with no more than a high school education. In
some speciﬁcations, refundable state EITCs are correlated with increased labor-force
participation, but the results are not statistically signiﬁcant. The minimum wage is found
to be correlated with increased probabilities that single women are out of the labor force
and in poverty. The results also indicate that welfare waivers are associated with
increased labor-force participation and reduced poverty, although waivers are found to be
less effective than expansions of the EITC.

Among women with children, the EITC subsidizes labor-force participation, with
the effect particularly large among younger women. It is possible that increased
participation will result in reduced school enrollments, especially among younger single

mothers. If such a trade-off occurred, it would represent a substitution of experience for

education, a substitution that will potentially affect long—term economic success. This
dissertation also uses data from the March CPS to investigate the impact of the EITC on
school enrollment among young, single women. Estimates obtained using a multinomial
logit model indicate that expansions of the EITC are associated with significant increases
in both employment and enrollment. The expansions of the EITC are associated with

signiﬁcant declines in young, single mothers being non-enrolled and non-employed.

C0pyright by
SCOTT B. DARRAGH
2002

To my wife Kendel and the memory of Bernard Scott Darragh

ACKNOWLEDGEMENTS

I would like to thank my committee chair, Professor Charles Ballard, for his
patience and diligence on this project. His assistance with the focus of the exposition has
been tremendous. I would also like to thank the other members of the committee;
Professor David Neurnark, Professor John Goddeeris, Professor Paul Menchik, and
Professor Dale Belman for providing me with helpful suggestions that enabled me to
better explore and explain this project.

My parents have always been supporters of my academic endeavors. Their love
has been a solid base from which this child could launch many pursuits. My only regret
in this pursuit is that my father is not here to see its completion.

No one has paid a greater price to see this project completed than my wife,
Kendel. Her persistent encouragement has often provided me with the necessary energy
to continue. I will always be grateful to her for her love and support.

“Of making many books there is no end, and much study is wearisome to the

ﬂesh” (Ecclesiastes 12:12b). Praise the Lord, this book is complete.

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix

LIST OF FIGURES ........................................................................................................... xi

INTRODUCTION ............................................................................................................... 1

CHAPTER 1

OVERVIEW OF THE EARNED INCOME TAX CREDIT (EITC) .................................. 3
INTRODUCTION ................................................................................................... 3
FEDERAL EITC ...................................................................................................... 7
COMPARISON OF THE EITC WITH OTHER PROGRAMS ............................ 15
STATE-LEVEL EITC ........................................................................................... 19

CHAPTER 2

LITERATURE ON LABOR SUPPLY AND THE EITC ................................................. 36
LABOR SUPPLY AND TAXES ........................................................................... 36
EITC AND LABOR SUPPLY ............................................................................... 38

CHAPTER 3

EFFECT OF THE EITC ON LABOR SUPPLY ............................................................... 45
OVERVIEW .......................................................................................................... 45
IDENTIFICATION ................................................................................................ 47
MODEL ................................................................................................................. 52
DATA .................................................................................................................... 57
ANALYSIS OF THE RESULTS ........................................................................... 60
CONCLUSION ...................................................................................................... 77

CHAPTER 4

EFFECT OF THE EITC ON POVERTY .......................................................................... 88
INTRODUCTION ................................................................................................. 88
EFFECT OF GENERAL CONTROLS ................................................................. 88
EFFECT OF THE EITC ........................................................................................ 91
COMPARISON WITH MINIMUM WAGES AND WELFARE WAIVERS ...... 98
ESTIMATES FOR 150 PERCENT OF POVERTY ........................................... 101

CHAPTER 5

COMPARISON TO PREVIOUS RESEARCH AND CONCLUSIONS ........................ 110
COMPARISON TO PREVIOUS RESEARCH ................................................... 110
CONCLUSION .................................................................................................... l 15

vii

CHAPTER 6

EFFECT OF THE EITC ON HUMAN-CAPITAL ACCUMULATION ........................ 118
INTRODUCTION ............................................................................................... 1 18
DESCRIPTION OF STUDY ............................................................................... 120
RELATED LITERATURE .................................................................................. 125
MODEL ............................................................................................................... 136
DATA .................................................................................................................. 140
PROBIT RESULTS ON PARTICIPATION ....................................................... 143
MULTINOMIAL LOGITS ON ENROLLMENT STATUS ............................... 149
ALTERNATIVE CALCULATIONS .................................................................. 154
COMPARISON WITH PRIOR LITERATURE ................................................. 157
CONCLUSIONS AND EXTENSIONS .............................................................. 160

APPENDIX ...................................................................................................................... 185

BIBLIOGRAPHY ............................................................................................................ 203

viii

LIST OF TABLES

CHAPTER 1
Table 1.1 History of Federal EITC ...................................................................... 34
Table 1.2 State Earned Income Credits ................................................................ 35
CHAPTER 3
Table 3.1 Sample Characteristics ......................................................................... 79
Table 3.2 Participation and Poverty Statistics Over Time ................................... 80
Table 3.3 Estimated Effects of the EITC on Labor Supply ............... 81
Table 3.4 Estimated EITC Elasticity of Real Earnings ........................................ 83
Table 3.5 Comparison of New York vs. US. Averages ...................................... 84
Table 3.6 Estimated Effects of the EITC on Labor Supply Using Lagged EITC
Controls ................................................................................................ 85
Table 3.7 Estimated Effects of Waivers, Welfare Reform, and the Minimum
Wage .................................................................................................... 86
Table 3.8 Estimated Effects of Waivers and the Minimum Wage with no EITC
Controls ................................................................................................ 87
CHAPTER 4
Table 4.1 Estimated Effects of the EITC on Poverty ......................................... 105
Table 4.2 Estimated Effects of the EITC on Poverty Using Lagged EITC
Controls .............................................................................................. 107
Table 4.3 Estimated Effects of the EITC on 150 Percent of Poverty ................ 108
CHAPTER 5
Table 5.1 Estimated Effects of the EITC on Labor Supply Using Sample of
Single Women Ages 18 — 44 ............................................................. 117
CHAPTER 6
Table 6.1 Employment and Enrollment for Single Women Ages 17 -— 21 ........ 163
Table 6.2 Employment and Enrollment for Single Women Ages 17 -— 24 ........ 164
Table 6.3 Summary of Minimum Wage Literature ............................................ 165
Table 6.4 Summary of EITC Literature ............................................................. 166
Table 6.5 Sample Summary Statistics ................................................................ 167
Table 6.6 Probits on Employment ...................................................................... 168
Table 6.7 Probits on Employment for Single Women Ages 17 — 21 ................. 169
Table 6.8 Probits on Employment for Single Women Ages 17 — 24 ................. 170
Table 6.9 Multinomial Logits on Employment and Enrollment for Single
Women Ages 17—21 ......................................................................... 171
Table 6.10 Partial Derivatives from Table 6.9 ..................................................... 172
Table 6.11 Multinomial Logits on Employment and Enrollment for Single
Women Ages 17 — 21 ......................................................................... 173
Table 6.12 Partial Derivatives from Table 6.11 ................................................... 174

ix

Table 6.13

Table 6.14
Table 6.15

Table 6.16
Table 6.17

Table 6.18
Table 6.19

Table 6.20
Table 6.21

Table 6.22

APPENDIX

Table A-1
Table A-2
Table A-3
Table A—4
Table A—5
Table A-6
Table A-7
Table A-8
Table A-9
Table A-lO
Table A-ll
Table A-12
Table A-13
Table A-14
Table A-15
Table A-16

Table A-17

Multinomial Logits on Employment and Enrollment for Single

Women Ages 17 — 21 ......................................................................... 175
Partial Derivatives from Table 6.13 ................................................... 176
Estimated Changes in Probabilities for Women with Two or More

Children .............................................................................................. 177

Estimated Changes in Probabilities for Women with One Child ...... 178
Multinomial Logits on Employment and Enrollment for Single

Women Ages 17 — 24 ......................................................................... 179
Partial Derivatives from Table 6.17 ................................................... 180
Multinomial Logits on Employment and Enrollment for Single
Women Ages 17 - 24 ......................................................................... 181
Partial Derivatives from Table 6.19 ................................................... 182
Multinomial Logits on Employment and Enrollment for Single
Women Ages 17 — 24 ......................................................................... 183
Partial Derivatives from Table 6.21 ................................................... 184
EITC and Federal Tax Statistics by State — 1998 .............................. 186
Unemployment Rate .......................................................................... 187
Poverty Rate ....................................................................................... 188
Labor Force Participation .................................................................. 189
AFDC/TANF Caseloads Per 1,000 People ........................................ 190
Probits Using Combined Federal and State Subsidy Rate ................. 191
Probits Using Separate Federal and State Subsidy Rate .................... 192
Probits Using Separate Federal and Refundable State Rate .............. 193
Probits Using Combined Federal and State Subsidy Rate with Low-
Education Subsample ......................................................................... 194
Probits Using Separate Federal and State Subsidy Rates with Low-
Education Subsample ......................................................................... 195
Probits Using Separate Federal and Reﬁindable State Rates with Low-
Education Subsample ......................................................................... 196
Probits Using Combined Federal and State Subsidy Rate with
Year/Child Interactions ...................................................................... 197
Probits Using Separate Federal and State Subsidy Rates with
Year/Child Interactions ...................................................................... 198
Probits Using Separate Federal and Refundable State Rates with
Year/Child Interactions ...................................................................... 199
Probits Using Combined Federal and State Subsidy Rate with
Year/Child Interactions and Low-Education Subsample ................... 200
Probits Using Separate Federal and State Subsidy Rates with
Year/Child Interactions and Low-Education Subsample ................... 201
Probits Using Separate Federal and Refundable State Rates with
Year/Child Interactions and Low-Education Subsample ................... 202

CHAPTER 1
Figure 1.1
Figure 1.2
Figure 1.3
Figure 1.4

LIST OF FIGURES

Effect of EITC on Budget Set .............................................................. 30
EITC’s Effect on After-Tax Income .................................................... 31
Effect of EITC on Equilibrium Wage .................................................. 32
Effect of Minimum Wage on Employment ......................................... 33

xi

INTRODUCTION

The earned income tax credit (EITC) has become the largest federal expenditure
program aimed at increasing the incomes of the working poor in the United States. The
credit began in 1975, but was almost ignored until the mid-19805. From 1986 to 1993,
the credit experienced three large expansions, which put the EITC at the forefront of US.
welfare policy.

The EITC is available to all low-income families with children. The credit
provides an initial wage subsidy that is designed to encourage non-working families to
enter the labor market. With increased earnings the credit is phased out, potentially
discouraging additional labor supply.

This dissertation investigates the impact of the earned income tax credit on labor
supply, poverty, and human-capital accumulation. Consistent with the primary focus of
the existing EITC literature, the study focuses on the impact of the EITC among single
mothers. Using data from the March CPS, a differences-in-differences framework is
employed to estimate the effectiveness of the EITC at raising labor-force participation
among single women ages 25 to 50. The estimates imply that the EITC is associated with
signiﬁcant increases in participation, especially among women with no more than a high
school education. The results provide strong evidence that the EITC is more effective at
increasing labor—force participation than either welfare waivers or increases in the
minimum wage.

The EITC is also found to be effective at reducing poverty among single mothers.

Once again, the impact is largest among women with lower levels of education. The

results are generally robust across a range of speciﬁcations, providing strong evidence
that the EITC is not only associated with increased labor supply, but also signiﬁcant
improvement in the economic well being of single mothers. The evidence indicates that
the EITC is more effective at reducing poverty than either the minimum wage or welfare
waivers.

In Chapter 6, the focus of the study turns to the impact of the EITC on human-
capital accumulation, speciﬁcally, what effect the EITC has had on the enrollment status
of young, single mothers. This investigation used a sample of single women drawn from
the March CPS from 1987 to 1996. By use of probit and multinomial logit techniques,
the study identiﬁes the impact of the EITC on the labor—market behavior and school
attendance for single mothers between the ages of 17 and 24.

The evidence indicates that expansions of the EITC are associated with large
increases in labor-force participation among young, single mothers. The results also
indicate that the increased labor supply does not reduce school enrollment. Rather, the
EITC is associated with both increased participation and enrollment.

This dissertation has six chapters and an appendix. Chapter 1 describes the EITC
in detail. Chapter 2 provides a review of the existing literature on taxes, labor supply,
and the EITC. Chapter 3 develops the empirical model for estimating the labor-supply
effects of the EITC and presents the results. Chapter 4 presents the empirical results for
the effect of the EITC on poverty. A comparison between the results in this study with
the existing literature is presented in Chapter 5. Chapter 6 contains the estimates for the
EITC’s effects on labor-market participation and school enrollment. The appendix

contains the detailed tables with the estimates discussed in Chapters 3 and 4.

CHAPTER 1
OVERVIEW OF THE EARNED INCOME TAX CREDIT (EITC)

1.1. INTRODUCTION

The federal earned income tax credit (EITC) has become the largest, non-health
related expenditure program in the United States targeted explicitly at improving the
welfare of the poor. Annual EITC expenditures are nearly equivalent to expenditures for
Temporary Assistance to Needy Families (TANF) and Food Stamps combined.1 The
program experienced two dramatic expansions in the 19905. These have expanded the
program from a relatively small $2 billion program in 19862 to a program with almost
$30.6 billion in expenditures in ﬁscal year 2001.3 The EIT C is a refundable income-tax
credit that can be used to offset federal income-tax liability. If the credit exceeds an
individual’s income-tax liability, the excess is paid as a refund. Most of the beneﬁts
distributed through the earned income tax credit are paid as income-tax refunds. The
US. Department of Treasury estimates that in ﬁscal year 2001, the program’s beneﬁts
will represent $4.8 billion in foregone income-tax revenues, and $25.8 billion in refund
payments.4

The program is designed to aid the working poor by providing cash assistance to
low-income individuals and families with eamed income. The credit applies to tax units
(either individuals or couples ﬁling a joint return) that have qualifying children living in
their home. Since 1993, low-income individuals aged 25 and over without children have

been eligible to claim a credit as well.

 

' FY 2002 Economic Outlook, Highlights from FY 1994 to FY 2001, FY 2002 Baseline Projections,
January 2001.

2 Holtzblatt, et a1. 1994.

3 Table 5-3, “Analytical Perspectives”, Budget of the United States, 2001.

The EITC’s growth reﬂects its effectiveness as a means of providing assistance to
the low-income population. The program has the highest participation rate by the eligible
population of all the programs targeting low-income groups, estimated at between 80 to
86%.5 This compares with estimated participation rates of 50 to 60 percent for the
Supplemental Security Income (SSI) program6 and 54 to 66 percent for Aid to Families
with Dependent Children.7

A number of factors combine to account for this. The integration of the EITC
with the federal income tax enables eligibility to be determined directly each year when
the return is prepared. For individuals who do not prepare their own return, a paid
income-tax preparer can easily determine whether a client qualiﬁes for the EITC. This
compares favorably with the lack of understanding regarding the eligibility standards for
other programs designed to assist those on public assistance, such as food stamps.8 Its
integration with the tax system also provides the EITC with an inconspicuous method of
distributing beneﬁts. Recipients receive beneﬁts without the stigma potentially
associated with receiving other forms of public assistance.

The growth of the EITC may also reﬂect a growing expectation among the US.
population that individuals able to work should work to contribute to their own support.
President Ronald Reagan championed the ﬁrst major expansion of the EITC precisely for
this reason.9 When President Clinton took ofﬁce in 1993, his initial State of the Union

address reinforced this objective. "The new direction I propose will make this solemn,

 

‘ Table 5-3, “Analytical Perspectives”, Budget of the United States, 2001.

‘ Scholz, 1994.

‘5 Haveman, 1987.

7 Blank and Ruggles, 1993, cited in Scholz, 1994.

8 For example, “Millions Eligible for Food Stamps Aren’t Applying” from The New York Times on the
Web, February 26, 2001. www.nytimes.com/2001/02/26/national/26FOOD.html.

9 Johnson, 1999.

simple commitment: by expanding the refundable earned income tax credit, we will
make history; we will reward the work of millions of working poor Americans by

realizing the principle that if you work 40 hours a week and you've got a child in the
house, you will no longer be in poverty."10

The EITC is designed to encourage work by increasing the total income received
by a participant above the income received in the labor market alone. The credit clearly
accomplishes this objective since, for any given number of hours worked, total income is
strictly greater than labor income for all workers eligible for the credit. These additional
beneﬁts from the credit will increase the resources available to a household. However, if
individuals make incremental decisions on labor supply, they are not concerned with the
total return to work, but rather with the return to the marginal hour of work, i.e., the net
wage for an additional hour of work. As discussed below, the EITC clearly encourages
entering the labor market. However, for those who already have earnings, the EITC’s
effect on additional labor supply is less clear.

The EITC could reduce poverty through a couple of channels. The credit raises a
household’s disposable income, for any ﬁxed amount of labor income. This will not
affect the measured poveny rates, though. EITC payments, almost 85 percent of which
are disbursed as tax refunds, are not included in income for the poverty statistics reported
by the US. Bureau of the Census. The credit either reduces income-tax liability or is

paid in a tax refund, and neither of these is included as a source of income in the general

poverty statistics.ll Over time, however, policies like the EITC that encourage labor

 

’0 Shapiro and Greenstein, 1993.

H For the small number of recipients who receive the credit through the advanced-payment option, the
EITC may be included in income. The EITC, while not included in the ofﬁcial poverty statistics, is
included as income in some of the broader measures of poverty.

supply may encourage the formation of human capital through work experience and
attachment to the labor market. Work experience may increase the probability that a
woman is employed in the future, and may reduce the probability that she and her family
are in poverty. Most directly, the EITC encourages labor supply, which results in higher
current labor income.

In this study I investigate two questions. Among women over age 24 with
children, is there evidence that the EITC is effective at encouraging labor supply,
especially when compared with other programs and policies designed or advocated to
support low-income individuals? This will involve comparisons with welfare beneﬁts,
variations in welfare policy, and minimum wages. The intent of this section is to
demonstrate whether the estimated impacts of the EITC on labor supply in this
population are similar to those reported in previous research by Eissa and Liebman
(1996), Meyer and Rosenbaum (2000 and 2001), Ellwood (2000), and Neumark and
Wascher (2001). The model used for this investigation is at least slightly different from
the techniques employed by each of the studies mentioned above.

With the impact of the EIT C on labor supply established, the investigation turns
to poverty. Speciﬁcally, what effect, if any, has the expansion of the EITC had on the
poverty status of single mothers? This is a relatively new question in relation to the
EITC, examined in only one other study (Neumark and Wascher (2001)). It is possible
that work subsidies like the EITC may increase labor-force participation, but still leave
many women and their children in poverty. The impact, if any, of the EITC on poverty
seems to be a fundamental question related to the credit’s effectiveness. This is the goal

behind the most recent expansion of the EITC, as described by President Clinton above.

While connecting work with assistance for low-income families may be important for
other reasons, it would seem that one of the underlying goals of any assistance program

should be to raise the recipients above poverty.

1.2. FEDERAL EITC

The earned income tax credit was enacted in 1975. When the credit was initially
adopted, it was designed to alleviate the burden of the Social Security payroll tax on
working parents.12 The program remained very small until the expansions of the 19905.
The EITC is a refundable income-tax credit, meaning that if the EIT C exceeds an
individual’s federal income-tax liability, the excess is refunded. In the tax calculation,
the EITC is treated in the same manner as a payment, such as withheld taxes.

Table 1.1 shows the history of the EITC from its enactment through tax year
2001. The rates in Table 1.1 ignore the supplemental credits, which were instituted in
1991 and continued until 1993. These provided extra beneﬁts for newborn children, as
well as for the purchase of health insurance for qualifying children.

The credit has three distinct regions based on the recipient's labor income: the
phase-in region, the cash-transfer region, and the phase-out region. ’3 For anyone outside
the labor market who would be eligible for the EITC, the credit increases the net wage
above the market wage. This increase can be quite generous, reaching 40 percent of
earned income in the wage-subsidy region following the full implementation of the 1993

expansion. For eligible individuals or families with more than one child, the phase-in

 

'2 For a thorough discussion of the legislative history of the EITC, see Ventry (2000).
'3 The analysis of the EITC presented here has beneﬁted from the treatment by Holtzblatt, et al. (1994),
Scholz (1994), and Browning (1995).

rates have nearly tripled from 1989 to 1996 (from 14 to 40 percent). The rates have also
risen sharply for those with one child (from 14 to 34 percent).

The cash-transfer region follows. While in this range of income, additional
earnings have no effect on the size of a recipient’s EITC. So, in this region, the EITC is
equivalent to a lump-sum transfer. The subsidy rate and the phase-in range determine the
EITC amount received by each recipient within this range.

At higher levels of income, the dollar amount of the credit is reduced by
additional labor income. This phase-out of beneﬁts lowers the net wage to a participant
below the market net wage, effectively taxing additional income. If labor-market
participants have a positive labor-supply elasticity, the credit will reduce the number of
hours eligible workers choose to work. So while the credit unambiguously raises the
total income of a worker with no change in hours worked, the credit may affect labor
supply and labor income. 14

As mentioned above, the EITC has undergone two major expansions in recent
years. During the period between the passage of the Tax Reform Act of 1986 and the
passage of the Omnibus Budget Reconciliation Act of 1990 (OBRA 1990), the phase-in
rate for the earned income tax credit was 14 percent and the phase—out rate was 10
percent. The passage of OBRA 1990 increased both the phase-in and phase-out rates,
created a slightly higher credit for families with more than one child, and provided
additional beneﬁts for new-born children and health insurance for children.

OBRA 1993 removed the requirement that a household must have a qualifying

child in order to receive an earned income credit. Individuals and married couples with

 

14 Some values of the labor-supply elasticity would be sufﬁcient to offset the effect of the EITC, resulting
in unchanged or reduced income.

earned income, total income below $10,710, and no children are eligible for a smaller
credit. The credit for recipients without children functions in the same manner as the
credit for families with children. The subsidy rate is exactly equal to the OASDHI
employee payroll tax rate (7.65 percent), and the tax rate in the phase-out range doubles
the payroll tax rate for this income group.

By 2001, the goal expressed by President Clinton was closer to reality.
Individuals and families with children and adjusted gross incomes below a certain limit
($32,121 in 2001 for a family with more than one child) were eligible for the credit. The
limits were lower for families with one child ($28,281) or no children ($10,710). An
individual working full-time at the federal minimum wage of$5. 15 with one qualifying
child would receive the maximum credit of $2,428. Assuming the individual works
2,000 hours per year and is ﬁling as the head of a household, the individual would have
no federal income-tax liability, and would receive the entire credit as additional income.
The $12,728 in gross income (including the EITC) would raise the two-person family
above the poverty threshold for 2001. The credit would increase take-home pay (gross
pay less the payroll tax) for those with earnings up to $11,940, before state income taxes.
This represents an increase in after-tax income of 25.5 percent. For families with two or
more children and one worker employed full-time at the minimum wage, total take-home
pay before state taxes would increase to $13,520, an increase of 42.1 percent due to the
earned income credit. Total gross income, including the employee portion of Social
Security and Medicare, for a family with two or more children and a full-time worker at
the minimum wage would increase to $14,308. That total income would raise a family of

three above the 2001 poverty threshold of $14,269.

A full-time worker working 2000 hours per year at the current minimum wage of
$5.15 would earn $10,300 from his or her employer. Assuming the individual is either
single or has a non-working spouse, the EITC would supplement that with an additional
$2,428 for the parent with one child, or $4,008 for the parent with two or more children.
This equates to a nominal wage of $6.46 for the parent with one child, or $7.32 for the
parent with two or more children, adjusting for the employee portion of the payroll tax.

For individuals not currently working, the EITC acts like a wage subsidy, raising
the net wage of the potential recipient. There is a constant subsidy rate in the phase-in
region, so the absolute size of the credit increases with increases in the recipient's earned
income. This is the region between 0 and Y; in Figure 1.1. In this region, the net wage
including the credit is higher (the budget set is steeper) than without the credit. This
wedge between the market wage and the net wage should increase labor—market
participation, since the return to work is increased and the labor cost of the worker to any
ﬁrm is unaffected.15 The subsidy rates have been increased by the recent changes in the
EITC. Families receive a subsidy rate of 34 percent with one child and 40 percent if they
have more than one child (see Table 1.1). This should have an unambiguously positive
effect on labor supply for those who are not currently in the labor force, if the
participation decision reﬂects a positive wage elasticity. There is no counteracting
income effect affecting the participation decision for those not currently in the labor
force, since labor supply is beginning at a comer solution with zero hours of work. State-

level credits, especially refundable credits, should enhance this effect.

 

'5 In general equilibrium, the increase in labor supply will lead to a reduced market wage, so the
employer’s labor cost will be reduced.

10

The subsidy’s effect on labor supply is theoretically ambiguous for those already
employed earning below Y1. A wage subsidy will increase labor supply through the
substitution effect, because the increased net wage will increase the cost of leisure,
leading individuals to consume relatively less leisure. However, the subsidy produces a
counteracting income effect. The additional income provided by the credit will increase
the individual’s income, enlarging the budget set, and thus leading to an increase in the
demand for all normal goods, including leisure. This should reduce labor supply. The
magnitudes of the substitution and income effects will determine whether individuals in
the phase-in range will increase or decrease their labor supply in response to changes in
the level of the credit. It is common for empirical studies to ﬁnd that the substitution
effect outweighs the income effect,16 but if the magnitude of the income effect increases
with income, this could change as higher levels of income remain in the phase-in region.

The cash-transfer region of the earned income tax credit acts as a strict cash grant.
This region is shown in Figure 1.1 between Y1 and Y2. Note that the two budget sets
representing the EITC (federal and state) are parallel to each other and the original budget
set between Y1 and Y2, reﬂecting the constant size of the cash grant. This region is for
incomes of between $10,000 and $13,100 for recipients with two or more children, with a
constant grant of $4,008 for individuals within this income region. The region begins at
lower levels for recipients with one child or no children, with lower total cash transfers
available for these groups as well. The cash-transfer region for recipients with one child

begins at $7,100 and ends at $13,100, while a recipient with no children receives the cash

 

'6 See Holtzblatt et al., for a summary.

11

grant with income of at least $4,750 and less than $5,950. The cash grants for recipients
with one child or no children are $2,428 and $364, respectively.

Recipients with incomes in the cash-grant range are only inﬂuenced by the
income effect. The size of the grant is independent of income in this range. Therefore, as
long as the recipient’s income level remains within the region’s boundaries, there is no
direct effect on the net wage for the recipient. Additional earnings within the cash-
transfer region will leave the amount of the transfer unchanged. The income effect
should have a negative impact on labor supply in this region. The additional disposable
income the recipient receives due to the grant should increase demand for all normal
goods, including leisure. Increased demand for leisure should reduce labor supply.

The ﬁnal region of the EITC is the phase-out region, where beneﬁts are reduced
when earnings increase. Three factors determine the impact the phase-out region will
have.17 The ﬁrst of these is the cash grant amount. The higher the beneﬁt, the greater
will be the phase-out or marginal tax rate in this region, with the other factors remaining
unchanged. The higher the tax rate, the greater the substitution effect away from labor
supply, since the net return to additional hours of work, i.e., the net wage, is reduced.

The second important factor is the length of the income interval chosen for the phase-out
of the credit. The level where the credit is zero is commonly called the break-even level
of income. The lower the break-even level, and thus the shorter the interval, all other
things equal, the higher will be the marginal tax rate required to phase out the credit. The

ﬁnal factor is this tax rate, also known as the beneﬁt-reduction rate. Only two of these

 

‘7 Browning, 1994.

12

factors are independent, i.e., once two of them are chosen, the third has been chosen by
default.

The phase-out range can be described as follows:

Beneﬁts = {1 - (Y - YCG) / (B - YCG)} x YG ,
where B is the break-even level of income, Y is the recipient's annual income, YCG is the
income level where the pure cash transfer ends and the phase-out region begins, and Y5
is the maximum EITC beneﬁt (or the income guarantee under a more traditional negative
income tax). The beneﬁt-reduction rate for the EITC is the change in beneﬁts for each
additional dollar of income, or, assuming the beneﬁt function is continuous,

dB /dY = - (B — YCG)" x Yo.

A credit of $3,000, which is going to be phased out evenly over the interval of
$10,000 to $25,000, will imply a tax rate of 20 percent. Lowering the tax rate will
require extending the credit to a higher income level or reducing the level of cash
assistance. The phase-out region is shown in Figure 1.1 between Y2 and B. After-tax
income increases by less than the wage for each hour of work, reﬂecting the phase-out
rate of EITC beneﬁts. This can be seen in the ﬁgure from the ﬂatter budget set between
Y2 and B.

Theoretically, the phase-out range has an unambiguously negative impact on
labor supply, assuming leisure is a normal good. Since the EITC increases the budget set,
i.e., it increases income for any positive number of hours worked as long as income is
below the break—even level, recipients will demand increased amounts of all normal
goods, including leisure. Even though the credit is getting smaller, it is possible the

income effect is getting larger as the recipient's total income increases, depending on the

13

impact of the level of income on the labor-supply decision. The phase-out rate also
imposes a wage tax on workers, thus lowering the net wage they receive. By the
substitution effect, this should reduce labor supply, since the reduction in the net wage
lowers the effective price of leisure. As shown in Table 1.1, the phase-out region begins
at income levels of $5,950 and $13,100, for eligible recipients with no children and one
or more children, respectively, in 2001. The phase-out region continues until income
reaches $10,710 for recipients with no children, until $28,281 for those with one child, or
until $32,121 for those with more than one child. The amounts of the cash grants and the
lengths of the phase-out regions imply phase-out rates of 7.65 percent for recipients with
no children, 15.98 percent for those with one child, and 21.06 percent for those with more
than one child, respectively.

The above description of the impact of the earned income credit has ignored the
interaction between additional income for the recipient and the amount of other forms of
government assistance for which the recipient is eligible. Many studies cite the high
marginal tax rates faced by participants when they participate in numerous programs such
as AFDC, Food Stamps, or Supplemental Security Income, as well as receiving an earned
income credit. When the beneﬁt-reduction rates of all of these programs are combined,
they make the marginal tax rate for some recipients of the EITC at least as high as it is for
any other group in the country.18 It is important to know that the earned income credit
has not traditionally reduced the recipient's eligibility for other programs such as Aid to
Families with Dependent Children (AFDC) and Food Stamps. The beneﬁt-reduction

rates of these programs are somewhat offset by the phase-in rate in the wage-subsidy

 

18 Holtzblatt, et al. (1994); Browning ( 1995), and Dickert, et al. (1994).

14

region of the EITC, but are accentuated by the beneﬁt-reduction rate of the EITC for
those still eligible.

In addition, extra income may make the recipient responsible for paying income
tax, both at the federal and state levels. For instance, a head of household claiming the
standard deduction with two dependent children would begin to pay federal income tax
with adjusted gross income of over $15,350 in 2001, which is well below the end of the
phase-out region for the EITC.19 While EITC beneﬁts are not subject to tax, labor
earnings over that threshold would be taxable. In Michigan, this worker would have
potentially been paying state income tax on all income above $9,900. Having income
subject to state and federal income tax, Social Security payroll tax, and the phase-out rate
of the EITC would imply a marginal tax rate of 42.91 percent. This compares to a
marginal tax rate of 43.3 percent for the highest tax bracket, which applies to those with
taxable incomes over $297,350 for taxpayers ﬁling as the head of a household for tax
year 2001. If high tax rates alone signiﬁcantly reduce labor supply, it should be reﬂected
by the group in the phase-out range of the EITC. The marginal tax rates would be even
higher if the recipients participate in the food-stamp program or other government

assistance programs, which have been ignored here.

1.3. COMPARISON OF THE EITC WITH OTHER PROGRAMS
The earned income credit is distinct from other methods of transferring income to
the poor in that it encourages work, assuming a positive labor-supply elasticity.

Speciﬁcally, the credit is only available to those with earned income. For those with little

 

'9 This ignores the child tax credit and the child and dependent care credit, which the taxpayer may be
eligible to claim.

15

or no earned income, the credit increases the return to additional labor. Traditional
welfare systems generally reduce beneﬁts when the recipient earns extra income, thus
lowering the return to work. For the EIT C, this occurs only in the phase-out range, in
which it is likely that the worker already has a signiﬁcant commitment to the labor
market. Most in—kind beneﬁt programs reduce the beneﬁts a participant may receive
when the participant increases his or her earned income, although small initial amounts of
earnings may be ignored for determining beneﬁt eligibility. Compared with these
programs, the EITC is distinct in that it rewards participants with extra earned income in
the subsidy range. Figure 1.2 presents a graphical example of the EITC increasing labor
supply over traditional welfare programs.

In Figure 1.2, the individual is seeking to maximize utility while choosing
between leisure and income from the labor market. Traditional welfare programs reduce
beneﬁts as the individual earns additional income.20 This is reﬂected by the beneﬁt-
reduction rate. The result is that the individual achieves a utility maximizing choice of
Y1 in labor income, resulting in utility U1. By contrast, the EITC initially subsidizes
labor supply, resulting in a utility-maximizing choice with higher after-tax income (Y2)
and less leisure (more labor supply).

A beneﬁt of encouraging contemporaneous labor supply may be longer-term
connections to the labor market for the affected population. As a worker gains
experience, the worker’s human-capital investment should generally increase. For
workers with low levels of educational attainment, this may consist of very basic skills,

such as reliable attendance habits and following directions. As the worker gains

l6

experience on the job, pay and/or hours will potentially increase, resulting in a higher
income and reducing the probability the worker and her family will be in poverty. This
potential for longer-term beneﬁts is a clear advantage of the EITC when compared to
other low-income assistance programs. The long-term beneﬁts of the EITC are not
examined in this paper. An evaluation of these beneﬁts would be an important
contribution to the literature on the EITC.

The credit also has certain advantages over increases in the minimum wage. In
the phase-in range, the EITC creates a wedge between the market wage and the net wage
paid to workers. There are increased incentives for workers in this range to participate in
the labor market, and for those already participating to work more. There is, however, no
additional cost for an employer for this wage subsidy. In fact, new workers entering the
labor market result in an outward shift in the labor-supply curve. This is shown
graphically in Figure 1.3. The additional workers drawn into the labor market due to the
EITC result in a shift in labor supply, from LS] to LS2. To the extent allowed by
minimum-wage laws, additional supply puts downward pressure on the equilibrium
wage. The market wage falls from W1 to W2. This would cause movement along the
labor-demand curve, encouraging employers to hire more workers. The result is higher
levels of employment due to lower market wages, and higher total labor-market income,
due to more workers being employed receiving both the market wage and the EITC.

Increases in the minimum wage, while also raising the net return to workers,
mandate an increase in labor costs for employers. This increase in the wage rate lowers

the quantity of labor demanded, and thus lowers employment. A higher minimum wage

 

2° A small amount of initial earnings is often allowed before beneﬁts are reduced. This allows for a utility-
maxirnizing solution that has less labor income than the one shown in Figure 1.2, but at a lower level of

17

would encourage employers to reduce their payrolls and substitute higher-skill workers
and advanced production techniques for low-skill workers. The decrease in employment
tends to offset the higher incomes some minimum-wage workers receive.

The effects of a minimum wage are shown graphically in Figure 1.4. The
minimum wage enacted at a higher rate than the market wage rate results in a reduction in
the quantity of hours worked, from Ho to H1 in the diagram. The higher wage results in
an increase in the quantity of labor supplied, so that LS1 hours of labor are offered.
However, the labor market is constrained by the demand side, with employers choosing
to employ fewer workers at the higher minimum wage.

The very groups that are targeted for aid may well be the groups that are injured
by the increase in the minimum wage. Lower-skill workers, for whom labor market
participation may be more important, are more likely to be displaced, and thus their long-
term prospects are adversely affected. The net effect on lower-skill workers will depend
on the elasticity of demand for labor. Neumark and Wascher (2002) demonstrate that the
elasticity of demand for minimum-wage workers may be near -1 since the disemployment
. effects of a minimum wage increase are concentrated among the workers affected by the
minimum wage. This occurs even with an overall elasticity of labor demand of -0.1 for
all workers in the age group. If job losses due to minimum wages are concentrated
among low-income families, it is possible (even likely) that increases in the minimum
wage will produce offsetting effects. The higher minimum wage will raise the incomes
of some families who remain employed with the higher minimum wage. Other families

who become unemployed or face reductions in hours worked will see their incomes fall.

 

utility than the choice made with the EITC.

l8

The empirical estimates of Neumark and Wascher indicate that minimum wages increase
the probability that poor families escape poverty, but also increase the probability that
previously non-poor families fall below the poverty line. The net impact of the offsetting
effects is insigniﬁcant, leading Neumark and Wascher to conclude that minimum wages
do not reduce poverty.

A second advantage the EITC enjoys over the minimum wage is its greater
concentration of beneﬁts in the target population. The EITC is completely concentrated
in ﬁling units with taxable income of no more than $31,152. Alternatively, the minimum
wage has no income qualiﬁcations, and applies to almost all workers. Burkhauser,
Couch, and Wittenberg (1996) report that at least one-third of those working at the
minimum wage were in families with incomes higher than double the poverty level.
While any annual measure of income may fail to measure accurately the economic well-
being of a taxpayer, it would seem the EITC is more concentrated among families that are
below median income. In addition, since the EITC is not available to individuals under
age 25 without children, it does not subsidize leaving school for most young people,
while higher minimum wages may encourage higher-skill students to leave school sooner
(see Neumark and Wascher, 1996). However, it is possible that the availability of the
EITC could induce young women with children to leave school, take a job, and cease

being a dependent.

1.4. STATE-LEVEL EITC
In recent years, the earned income credit has become an attractive policy tool for

states as well. By 2000, 14 states and the District of Columbia had their own credits in

19

operation. Ten states have adopted new credits since 1990.21 In general, the state-level
credits use federal eligibility criteria, and their beneﬁts are determined as a percentage of
the federal credit. A reﬁrndable state EITC adds to the incentive effects for participation
of the federal credit. A refundable state credit also provides additional income support
for low-income families.

The current federal subsidy rate for families with two or more children is 40
percent. Since most state EITCs operate as a percentage of the federal credit, a
refundable state credit of 10 percent increases that rate to 44 percent. Refundable credits
have been adopted by Colorado, Kansas, Maryland, Massachusetts, Minnesota, New
Jersey, New York, Vermont, Wisconsin, and the District of Columbia. However, a
number of states have adopted nonrefundable credits. A nonrefundable credit has limited
incentive effects, because any credit in excess of tax liability goes unused.
Nonrefundable credits have been enacted by Illinois, Iowa, Maine, Oregon, and Rhode
. Island. Maryland has both a refundable and a nonrefundable option, and taxpayers may
choose the option that provides the largest beneﬁt.

Wisconsin can be used as an example to compare reﬁrndable and nonrefundable
credits. With the existing reﬁindable Wisconsin credit, a single mother with two children

and earnings of $1 1,635 would receive a Wisconsin earned income credit of $561 for

 

2' Rhode Island (1975), Maryland (1987), Vermont (1988) and Wisconsin (1989) were enacted prior to
1990. Iowa (1990), Minnesota (1991), New York (1994), Oregon (1997), Massachusetts (1997), Kansas
(1998), Colorado (1999), Illinois (2000), Maine (2000), New Jersey (2000), and the District of Columbia
(2000) have all adopted new credits, and Maryland (1998) revised their credit to grant a refundable option.
A 15'h state, Indiana, adopted an earned income credit that bears little resemblance to the other state-level
credits. The Indiana credit will be discussed later, but is not considered an earned income credit in the
above discussion.

20

2001.22 If the Wisconsin credit were to be nonrefundable, this single mother would
receive no beneﬁt from the nonrefundable state credit with an income of $1 1,635. Under
the 2001 Wisconsin income-tax law, the standard deduction, allowable personal
exemptions, and renter’s property-tax credit would eliminate all tax liability for this
hypothetical taxpayer. This analysis assumes this hypothetical mother pays $500 per
month in rent. So a nonrefundable credit would provide no effective beneﬁt for this
mother unless she earns more than $11,635 in 2001. With an income of $1 1,635, the
hypothetical family described above has no income—tax liability, making a nonrefundable
credit of no value.

State income taxes generally have exemptions and/or standard deductions that
provide an income threshold below which tax ﬁlers pay no income tax. A nonrefundable
credit will generally raise that threshold. By contrast, a refundable credit actually adds to
the federal subsidy, providing an additional incentive to a nonworker to begin working.
The excess earned income credit above any income-tax liability is reftmded to the
worker.

The federal earned income credit is often heralded as an incentive program
designed to encourage labor supply. In the phase—in region, the credit increases the net-
of-tax wage above the market wage, thus increasing the reward for market labor. This
should increase labor market participation. A state-level credit would increase this
subsidy even more, so it would be proper to adopt a state-level credit in the hopes of

increasing labor-market participation.

 

22 The Wisconsin credit for families with two children is 14 percent of the federal credit. More detail on
the Wisconsin credit will be given below.

21

All state earned-income credits are not created equal. The size of the state affects
the absolute amount a state would have to pay for any particular credit. For 1998, a 10-
percent credit in California would have cost $380 million, while a 10—percent credit in
Delaware would cost $8 million.23 However, states also differ in the relative size of the
EITC credit population. On average, 20 percent or more of federal income-tax returns
ﬁled from southern states claim an EITC, while the national average is only 15.5 percent.
In 1998, 29.8 percent of all returns ﬁled from Mississippi claimed an earned income tax
credit. States in the upper Midwest have smaller EITC populations, with 10-12 percent
of federal returns ﬁled by residents of these states claiming a credit. With states
generally basing their credits on a percentage of the federal credit, variation in the
proportion of taxpayers claiming the federal credit will mean that a credit of any given
percentage results in different relative expenditures for various states. A state with a low
proportion of claimants will ﬁnd its credit inexpensive relative to a state with a higher
proportion of claimants. This may in part explain the generous credit rates in states like
Minnesota and Wisconsin, and the fact that state-level EITCs have not been adopted in
the South. Appendix Table A-1 shows the distribution of credits for 1998 by state.

A brief sketch of the history and framework of the existing state-level EITCs is

given below. Existing state credits are also summarized in Table 1.2.

Rhode Island
The Rhode Island credit is the oldest state-level credit. The Rhode Island credit is

embedded within the state’s income-tax calculation, and not separately reported by

 

23 Data on federal EITC payments by state are taken from the 1998 Statistics of Income, Internal Revenue
Service.

22

taxpayers. The Rhode Island income tax is based on 26.5 percent of the federal income
tax after certain federal credits, including the earned income credit, so an extra $1 in
federal credit reduces the tax liability of a Rhode Island resident by $0.265. This credit is
nonrefundable, so if the net federal income-tax liability is negative due to refundable
credits, the Rhode Island tax liability is zero. This credit has been in place since the

beginning of the federal credit in 1975.

Wisconsin

The Wisconsin credit is unique in that the percentage of the federal credit an
eligible taxpayer may claim depends on the number of qualifying children in the
household. Wisconsin’s experience with an earned income credit began in 1984 as a
nonrefundable credit. This credit was then eliminated following the 1985 tax year. In
1989 Wisconsin adopted a new reﬁrndable credit that has been modiﬁed numerous times
in the years following. Since its return, the credit has varied in size based on the number

of qualifying children. Currently the state credit works as follows:

 

 

Number of qualifying children Percentage of the federal credit
1 4%
2 14%
3 or more 43%

23

Maryland

Maryland was the third state to adopt an earned income credit, in 1987. The
credit was equal to 50 percent of the federal credit and was nonreﬁrndable. In 1998,
Maryland added a refundable altemative. The new refundable portion is 10 percent of
the federal credit, and taxpayers may choose the alternative offering them the best

advantage. The refundable credit percentage increased to 15 percent for tax year 2001.

Vermont

Vermont has a refundable credit equal to 25 percent of the federal credit. The
credit has varied in its percentage since its inception in 1988. In 1988 and 1989, the
credit equaled the current 25 percent rate. From 1990 through 1992, the credit was 28
percent, and in 1993 it was 31 percent. In 1994 the credit was returned to its current 25

percent rate.

Iowa
The Iowa credit began in 1990 as S-percent nonrefundable credit. The credit was
increased to 6.5 percent of the federal credit in 1991. This was the smallest among the

state-level credits prior to 1997.

Minnesota

Minnesota’s credit was adopted in 1991 as a refundable credit equal to 15 percent

of the federal credit. In 1998, the credit was revamped. An eligible taxpayer without a

24

qualifying child receives a Minnesota credit equal to 15 percent of the federal credit. The
credits for taxpayers with children have a tiered approach. There is an initial subsidy
rate, a zero rate, a higher subsidy rate, a zero rate, and then the credit phases out.

Minnesota’s approach is unique among states with EITCs.

New York
New York created a refundable credit in 1994 that was phased in over a 4-year
period. In 1994, the credit was 7.5% of the federal credit. It increased to 10% in 1995,

15% in 1996, and 20% after 1996. The credit increased to 25% in 2001.

Massachusetts
Beginning in 1997, Massachusetts adopted a 10% refundable credit. The credit

rate increased to 15% in 2001.

Kansas

Beginning in 1998, Kansas adopted a 10% refundable credit.

Oregon

Beginning in 1997, Oregon adopted a 5% nonrefundable credit.

Colorado

In 1999, Colorado adopted a refundable credit equal to 8.5% of the federal credit.

The continuance of the credit is dependent upon Colorado maintaining a budget surplus.

25

Indiana

Beginning in 1999, Indiana created a different kind of earned income credit. The
credit does not work like other piggy-back credits. The credit applies to households with
, total federal income of $12,000 or less. The Indiana credit is equal to the 3.4 percent of
the difference between $12,000 and total federal income. This means the Indiana credit
operates like a cash grant of $408 (3.4 percent of $12,000) and has a beneﬁt-reduction
rate, i.e., an income tax rate, of 3.4 percent. This particular version of the earned income
credit does not feature any wage subsidy, unlike the federal credit or the credits adopted
by most states. It would seem that this program is designed to provide income
maintenance, since the program imposes a tax on additional earnings for low-income
workers.

Beginning in tax year 2000, 3 additional states and the District of Columbia added
earned income tax credits. New Jersey adopted a 10% refundable credit that will increase
to 20% by 2003. The District of Columbia has also adopted a 10% refundable credit.
Illinois and Maine began offering 5% nonrefundable credits.

A quick look at some summary statistics presented in Tables A-2 through A-5
would seem to indicate that the state-level earned income credits, with the exception of
New York and the District of Columbia, were not adopted by states experiencing acute
labor-market difﬁculties. As shown in Table A-2, in the year before adoption, the
unemployment rate in states that enacted their own credit was 0.6 percentage points
below the national average. The unemployment rate would be 0.9 percentage points
below the national average if New York and the District of Columbia were excluded.

The numbers are very similar if the annual average for the two prior years is considered.

26

A similar pattern emerges when considering poverty rates, as shown in Table A-3.
States that enacted EITCs averaged 2.3 percentage points less poverty than the national
average in the year prior to adoption. That number grows to 3.1 percentage points when
New York and the District of Columbia are excluded.

As can be seen in Tables A-4 and A-5, a similar situation exists for labor—force
participation and AF DC/TANF caseloads. States adopting credits were an average of 2.5
percentage points above the national average in labor-market participation. Excluding
New York and the District of Columbia would increase that number to 3.3 percentage
points. In the year before the state adopted an EITC, the number of AFDC/TANF cases
per 1,000 residents was 0.5 higher than the national average. However, that number
drops to —2.1 per 1,000 residents when New York and DC are excluded.

These four measures of economic well-being indicate that the states adopting
credits were, with the two notable exceptions, not suffering from dire economic times
relative to the US. as a whole. That does not exclude the possibility that there may be
regional differences in the tolerance of unemployment, poverty, and active welfare
policies. Thus, earned income credits may have been adopted to provide additional
resources for low-income working families, or to balance tax relief across income groups,
but it would not seem they were adopted to address serious labor-market problems. New
York and the District of Columbia may be exceptions. Both areas were characterized by
relatively high unemployment, poverty, and large proportions of their population on
either AFDC or TANF, while having relatively low labor-market participation. In these
areas, an EITC may have been viewed as a way to encourage work and reduce the need

for future public support. The credit will also provide a measure of support to low-

27

income workers, helping to alleviate poverty.

From this simple examination of the more recently created earned income credits,
it can be observed that New York may have adopted an eamed-income credit for different
reasons than other states. The high unemployment and poverty rates, as well as the low
participation rate in New York, may indicate that the New York EITC was viewed as a
policy tool to improve labor-market performance. Whatever the motivation, it is clear
that the situation in New York is different from the situation in the other states that have
adopted an EITC.

The assumption used throughout this paper is that the adoption of state-level
earned income credits is exogenous. A potentially interesting extension would be to relax
that assumption and consider the factors that inﬂuence states to adopt their own EITCs.
An endogenous state-policy regime may provide additional insight into the effectiveness
of EITCs. It is possible that the states that adopt EITCs have strong labor-market
performance prior to the enactment of the EITC. If state EITCs are assumed to increase
labor-market participation, states with high levels of participation, especially among
single mothers, might not receive as strong an increase in participation from the new
EITC as would states with lower participation. This would potentially bias the measured
effect of a state EITC toward zero. However, if EITCs are enacted by states with weak
labor-market performance, the measured effect of a state EITC would be biased upward.

The above discussion notwithstanding, for the purposes of this study, state-level
credits are assumed to be exogenous policy regimes, so that the natural experiment
approach is valid. State-level credits potentially improve the identiﬁcation of the effects

of EITCs on labor supply and poverty. The variance across states in subsidy and tax rates

28

associated with the various credits may provide more precise controls and better

identiﬁcation of the EITC’s overall impact.

29

 

Total Income after
Credits

 

FIGURE 1.1

Effect of EITC on Budget Set

    
   
 

 

 

 

State + Federal EITC _-__ /
Federal EITC
I,” » Original Budget Set
Y1 Y2 B
Earned
Income

30

FIGURE 1.2
EITC’s Effect on After-Tax Income

Leisure

 

 

    

Traditional \
Welfare
Program

 

 

 

 

Y1 Y2 After-tax income

31

FIGURE 1.3

Effect of EITC on Equilibrium Wage

Wage Rate

L82

L81

W1

p-——_____---

LD

 

"U'UIU'I'"IIUU""'E'-"""l"-""---'. "'l

/

l

 

W2

 

 

Hours

H2

32

FIGURE 1.4
Effect of Minimum Wage on Employment

Nominal
Wage

Labor

Demand Labor

Supply

Minimum Wage

 

 

Equilibrium Wage

 

 

 

 

 

 

Hours

33

Calendar
Year

1975-78
1979-84
1985-86
1987
1988
1989
1990
1991
One child
Two children
1992
One child
Two children
1993
One child
Two children
1994
No children
One child
Two children
1995
No children
One child
Two children
1996
No children
One child
Two children
1997
No children
One child
Two children
1998
No children
One child
Two children
1999
No children
One child
Two children
2000
No children
One child
Two children
2001
No children
One child
Two children

Table 1.1

History of Federal EITC

Credit Maximum Phaseout

Rate (%) Credit Rate (%)
10.00 $400 10.00
10.00 $500 12.50
11.00 $550 12.22
14.00 $851 10.00
14.00 $874 10.00
14.00 $910 10.00
14.00 $953 10.00
16.70 $1,192 11.93
17.30 $1,235 12.36
17.60 $1,324 12.57
18.40 $1,384 13.14
18.50 $1,434 13.21
19.50 $1,511 13.93
7.65 $306 7.65
26.30 $2,038 15.98
30.00 $2,525 17.67
7.65 $314 7.65
34.00 $2,094 15.98
36.00 $3,110 20.22
7.65 $323 7.65
34.00 $2,152 15.98
40.00 $3,556 21.06
7.65 $332 7.65
34.00 $2,210 15.98
40.00 $3,656 21.06
7.65 $341 7.65
34.00 $2,271 15.98
40.00 $3,756 21.06
7.65 $347 7.65
34.00 $2,312 15.98
40.00 $3,816 21.06
7.65 $353 7.65
34.00 $2,353 15.98
40.00 $3,888 21.06
7.65 $364 7.65
34.00 $2,428 15.98
40.00 $4,008 21 .06

Phaseout
Income

$4,000
$6,000
$6,500
$6,920
$9,840
$10,240
$10,730

$11,250
$11,250

$11,840
$11,840

$12,200
$12,200

$5,000
$1 1,000
$1 1,000

$5,130
$11,290
$11,290

$5,280
$11,610
$11,610

$5,430
$11,930
$11,930

$5,570
$12,260
$ 12,260

$5,670
$12,460
$12,460

$5,800
$12,700
$12,700

$5,950
$13,100
$13,100

Breakeven
Income

$8,000
$10,000
$1 1,000
$15,432
$18,576
$19,340
$20,264

$21,250
$21,250

$22,370
$22,370

$23,050
$23,050

$9,000
$23,755
$25,296

$9,230
$24,396
$26,673

$9,500
$25,078
$28,495

$9,770
$25,750
$29,290

$10,030
$26,473
$30,095

$10,200
$26,928
$30,580

$10,380
$27,413
$31,152

$10,710
$28,281
$32,121

Source: The 2000 Green Book, Ways and Means Committee, US. House of Representatives.

2001 information from Instructions for Form 1040, Internal Revenue Service.

34

Table 1.2
State Earned Income Credits

State Year Enacted Refundable Credit Rate - 2001
Rhode Island 1975 No 26.5%
Wisconsin 1984, 1989 Yes 1 child — 4%

2 children —- 14%
3 or more — 43%
Maryland 1987 1987-97 — No Nonrefundable — 50%

Since 1998 — Both Refundable - 15%

Vermont 1988 Yes 25%
Iowa 1990 No 6.5%
Minnesota 1991 Yes Variable
New York 1994 Yes 25%
Massachusetts 1 997 Yes 1 5%
Oregon 1997 No 5%
Kansas 1998 Yes 10%
Colorado 1999 Yes 8.5%
New Jersey 2000 Yes 10%
District of Columbia 2000 Yes 10%
Illinois 2000 No 5%
Maine 2000 No 5%

35

CHAPTER 2

LITERATURE ON LABOR SUPPLY AND THE EITC

2.1. LABOR SUPPLY AND TAXES

Labor supply among women has increased in every age group in the period
following World War II. From 1980 to 1998, participation increased by 8.4 percentage
points among all women.24 Double-di git percentage-point gains occurred among women
25 to 54. The smallest gains were among women 16 to 24, which is possibly due to
longer school enrollment. College enrollment among women grew from 1980 to 1998 by
43.3 percent, with an additional 2.6 million women enrolled for degree credit. College
enrollment among men over the same period grew 27.8 percent.25 The overall increases
in the labor supply of women have helped to spur research interest, both to describe the
changing labor-market behavior of women and to assess the effects of various policies on
the labor supply of women.

A deep literature exists on the overall effects of taxation, speciﬁcally income
taxation, on labor supply. This literature mainly deals with the substitution and income
effects of taxes on labor income, and provides intuition on the ﬁndings discussed later.

Numerous previous studies have found varying degrees of response in labor
supply to tax changes. T riest (1990) ﬁnds that, generally, changes in the net wage have
almost no impact on the labor supply of men. Small estimated labor-supply elasticities
(near zero) for men are common throughout the labor-supply literature. Estimated labor-

supply elasticities for women are shown by Triest to depend on the method used for

 

2‘ Fullerton (1999).
25 Data from the US. Census Bureau, Current Population Reports, P20-521.

36

estimation. In particular, conditional on participation in the labor force, the elasticities
for women are small, but the participation decision seems quite elastic. This would seem
to be good news, especially for those hoping to draw individuals into the labor market
with the high wage-subsidy rates in the phase-in portion of the EITC.

In Hausman's survey (1985), the weight of the evidence seems to indicate that
low-income individuals have a greater response to tax law changes than do other income
groups. Hausman's survey deals mainly with the results of the negative-income-tax
(NIT) experiments conducted in the 1970's. In these studies, the expected reduction in
labor supply to the NIT was observed. An assistance program with a NIT or beneﬁt-
reduction rate will have both the income and substitution effects working to reduce labor
supply. The NIT experiments consistently observed this empirically, both for men and
women.

Mofﬁtt's (1992) survey of studies does not address the EITC or the income tax
speciﬁcally, but rather looks at the other components of the US. welfare system. Many
of these programs have NIT features, which reduce the beneﬁts received by recipients as
their incomes increase. Mofﬁtt ﬁnds that, because changes in beneﬁt-reduction rates
cause changes in the population eligible for beneﬁts by changing income cutoffs, there is
no signiﬁcant effect on overall labor supply from rate changes. By raising the break-even
level of income, the beneﬁt-reduction rate is reduced, but individuals formerly ineligible
for beneﬁts now become eligible. Those who are newly eligible face an increased tax
rate because of the negative income tax structure of the program. Mofﬁtt reports that
most studies ﬁnd that these effects cancel each other. In fact, he reports, "The failure [of]

the negative income tax to provide work incentives has also led some analysts to

37

conclude that work and welfare are inevitably essentially mutually exclusive alternatives.
. .. (pg. 41)" The potential implication for the EITC is that while increasing beneﬁts,
such as was done in 1990 or 1993, may increase labor-market participation, the increased
beneﬁt-reduction rates or expansion of the population subject to the beneﬁt-reduction rate
will result in offsetting decreases in labor supply.

Mroz (1987) ﬁnds that the labor-supply behavior of married working women is
similar to that of prime-age males. The models that produce large income or substitution
effects require economic and statistical assumptions that are rejected by Mroz's
speciﬁcation tests. If married women, in fact, exhibit labor-supply behavior similar to
males, it would seem reasonable to expect similar behavior from female household heads,
the focus group of this study. The success of the recent expansions of the EITC will
likely be judged based on the impact these expansions have had on labor supply. If, as a
result of increasing the wage-subsidy rate and thus the size of beneﬁts, a larger
proportion of the low-income population is drawn into the labor market, the expansions
will be viewed as successful to some degree. If the result is to reduce the number of
hours worked signiﬁcantly by those already in the labor force, the expansions will

undoubtedly be viewed less favorably.

2.2. EITC AND LABOR SUPPLY

Dickert, Houser, and Scholz (1995) complete a microsimulation of the tax rates
faced by low-income households. By comparing the effective tax rates faced by different
households, the study projects the expected impact the expanded EITC will have on labor

supply. The simulation is quite detailed, and attempts to model the interaction between

38

the labor market, federal and state tax policy, the EITC expansion, and other government
assistance programs. Their projections would seem to indicate that the new expansions
of the EITC would lower effective tax rates, due to the expanded phase-in range and
higher subsidy rate. Lower tax rates should lead to increased labor supply and earnings
for low-income families.

Browning (1995) uses estimates of wage and income elasticities to estimate the
potential labor-supply effects of the EITC expansion resulting from OBRA 1993. These
estimated elasticities are then used to project the impact the expansion of the EITC would
have on households in the different EIT C ranges. The most striking ﬁnding was the
decline in disposable income projected to occur among households in the higher end of
the phase-out range. This implies that the decline in labor supply among these
households is so great that labor income would decline by enough to offset the higher
EITC amount. The estimates imply that the net beneﬁt to households in the phase-out
range would be signiﬁcantly less than the cost of providing the beneﬁt. However,
Browning’s estimates indicate that the EITC expansion will increase the disposable
income of families in the phase-in and cash-grant ranges.

The simulation results above are based on previously estimated elasticities. The
responsiveness of non-participants may vary over time, so the current group of non-labor-
market participants might not be as responsive, on average, to increased wage subsidies
as were those in the past. The pool outside the labor market may lack job skills which,
when combined with minimum wage laws, may seriously restrict opportunities for these

prospective workers. These workers also may not be strongly attached to the job market.

39

Some may be involved in underground activities, which could lessen the incentive effects
of the EITC.

Eissa (1995) looks at changes in the labor supply of married women in response
to the Tax Reform Act of 1986 (TRA). The focus of this study is on women in the upper
25% of the income distribution. The estimated labor-supply elasticity for married women
in this study is between 0.6 and 1. It is similar to other ﬁndings, but the composition of
the results are different. In particular, the estimated elasticity is approximately half due
to increased participation and half due to increased hours. This differs from the Triest
study above, in that he found that the hours elasticity was almost zero, while the
participation elasticity was the driving force behind the overall elasticity. The Eissa
study on high-income women is applicable only in the sense that it provides additional
evidence of labor supply responses to changes in labor-market incentives. The TRA did
not impose a phase-out rate, so the effect was only to increase overall labor supply. The
study is of limited applicability to the work presented here because the sample population
used by Eissa was high-income, married women. These women are not eligible for the
EITC, nor are their incentives affected by changes in welfare policy.

Eissa and Liebman (1996) examine the effects of the changes in the earned
income tax credit contained in the TRA on female heads of households which include
children. They ﬁnd that the expansion of the EITC made by the TRA led to an increase
of 1.9 to 2.8 percentage points in the labor-force participation of single women with
children, relative to single women without children. Eissa and Liebman did not ﬁnd any
signiﬁcant change in hours worked. This is the kind of response that is expected by

policymakers who advocate expanding the EITC, although the recent expansions of the

40

, EITC were larger in scope and more publicized than the earlier changes, and may have
different impacts.

The study by Eissa and Liebman was the ﬁrst to evaluate the EITC using
econometric techniques. Previous studies had focused on simulating the outcomes of an
expansion of the credit. In recent years, additional studies on the EITC have emerged.
The consistent ﬁnding throughout this literature is that expansions of the earned income
tax credit have been associated with increases in the labor supply of single women.
Meyer and Rosenbaum (2000, 2001) have shown in a pair of studies that recent federal
expansions of the EITC were effective at raising the labor-market participation of single
mothers. Their methodology attempts to incorporate various changes in the EITC,
welfare policies, and state-level policies (including EITCs) in a structural model of the
work decision, to glean the effects of these policies on the labor-supply decisions of
single women. The 2001 study ﬁnds that the EITC is the most signiﬁcant factor
accounting for the increase in the labor-force participation of single mothers. They ﬁnd
that welfare waivers also provided a positive inﬂuence on participation.

The 2000 paper makes large-sample comparisons, using samples drawn from the
March Current Population Survey (CPS) and the CPS Outgoing Rotation Group. The
changes in labor-force participation for the group most affected by welfare policy and
EITCs, single women with children, were compared with changes to other groups (single
women without children, married women with children, and black men). Employment
for single mothers, especially mothers with young children and those with multiple

children, rose signiﬁcantly more than employment in the comparison groups.

41

Ellwood (2000) focuses on the differential incentives created by the EITC and
other policy variables. Among the differences investigated in Ellwood’s paper are
changes in tax policy over time, the erosion of real welfare beneﬁts, and the
aggressiveness of state policies in attempting to move individuals from welfare into
work. Ellwood’s results suggest that, while the EIT C increases the labor-market
participation of single mothers, there is a small decrease in the participation of married
mothers.

Ellwood also considers the EIT C’s potential impact on marriage. Since low-
income taxpayers are less likely to have itemized deductions, it is almost certain that two
working individuals whose combined income falls within the EITC range would face a
marriage penalty.26 Since the standard deduction for married couples is lower than either
double the standard deduction for single ﬁlers, or the sum of the single and head-of-
household standard deductions, low-income households will likely pay more tax as a
married unit than as two single units. The EITC may exacerbate a marriage penalty on
low-income couples that decide to marry. Two workers who are employed at least half-
time for the tax year and EITC-eligible, would ahnost certainly be in the phase-out range
of the credit if they were to ﬁle jointly. Thus, their income would be taxed at high
marginal tax rates. In addition, a marriage between an EITC recipient and another
worker may result in the EITC amount being reduced, or even being eliminated. This is
because the couple together may have an income that is further in the phase-out range or

even outside EITC eligibility. This could lead to reduced marriage and increased

 

26 The following discussion on marriage penalties focuses on low-income taxpayers. Marriage penalties
on higher income taxpayers are primarily due to the tax rate structure. The higher tax rate brackets for
married couples begin at income amounts that are less than double the amounts for single taxpayers.

42

cohabitation due to expansions of the EITC. However, the Ellwood results suggest that
the EITC does not affect the decision to marry.

Eissa and Hoynes (1998) ﬁnd that the EITC reduces the probability a married
woman participates in the labor market. They also ﬁnd evidence that participants already
in the labor market reduced their supply due to the expanded EITC and the higher
beneﬁt-reduction rate. The reductions range from ~06 to —2 percent. This is consistent
with higher marginal tax rates discouraging participation, and generally consistent with
the projections of Browning discussed above. The effect on married couples with
children may be stronger than the effect on single mothers, because a married woman’s
earnings are more sensitive to tax policy, since the family has two potential workers. The
higher tax rates may encourage more substitution to household production.

The policy advocated most commonly to aid the working poor, especially outside
of academic economics, is undoubtedly raising the minimum wage. However, most
papers discussing the EITC or welfare reform omit references to it. One study
incorporating the EITC and minimum wages was from Neumark and Wascher (2001).

Neumark and Wascher was the ﬁrst paper to investigate empirically the impact of
the EITC on poverty status. This study matched CPS households over consecutive years
and found that changes in EITCs, especially those enacted by states, were a more
powerful tool for raising the incomes of poor families than minimum wages. This is
consistent with government wage subsidies being able to raise the return to labor without
distorting the labor market by imposing a legal wage ﬂoor. The ﬁnding that state EITCs
have a stronger effect than the federal EITC is different than one might expect

theoretically, although the structure of the model used by Neumark and Wascher provides

43

for better identiﬁcation of the impact of state EITCs.27 There is no particular reason that
a federal or state EITC should be more effective if both provide the same incentives and
are administered in a similar manner. Given that the federal EITC provides a larger wage
subsidy to encourage labor-force participation, it would be expected that the federal EITC
would have a more signiﬁcant impact than state EITCs.

The one difﬁculty with the study framework is that households only have two
years to observe the new incentives and respond. Since changes in tax policy may
require longer lags than minimum-wage laws, it might be expected that the full
minimum-wage effect might be observed within the two-year period, while part of the
EITC effect might not occur until after the second year. This would tend to reduce the
observed impact of EITCs. The results discussed in Chapter 3 support the hypothesis that
the effect of the EITC does occur with a lag. Thus it may be that Neumark and

Wascher’s results, while favorable to EITCs, understate the ﬁnal impact.

 

27 In fact, for a number of the speciﬁcations reported by Neumark and Wascher, only the state EITC is
identiﬁed.

44

CHAPTER 3

EFFECT OF THE EITC ON LABOR SUPPLY

3.1. OVERVIEW

The EITC clearly changes the incentives facing low-income individuals in the
labor market. The goal for this study is to evaluate how those incentives have affected
the labor-market participation and poverty status of single mothers. The quote referenced
at the beginning placed two objectives at the heart of US. welfare policy; getting people
to work and then ensuring that low-income working families are able to escape poverty.
It is intended that this study may help evaluate the success of these efforts. This study is
similar to studies performed by Eissa and Liebman (1996), Meyer and Rosenbaum (2000
and 2001), and Ellwood (2000). The focus here is on identifying all of the policy
variables used to target low-income families, including the EITC, minimum wage,
welfare beneﬁts, and welfare waivers such as term limits and work requirements.

The above studies focus primarin on the impact of the EITC on labor supply,
especially whether single mothers participate in the labor market. In addition to
participation, these studies also investigate the potential effects of the EITC on other
economic and social outcomes. Eissa and Liebman investigate whether the EITC is
associated with reduced hours of work for those single mothers already in the labor force.
In addition to labor-force participation, Meyer and Rosenbaum (2001) investigate the
effects of the EITC on the probability a single mother receives welfare beneﬁts. Ellwood
includes an analysis of the effects of the EITC on labor supply of low-wage married

women and the probability of marriage compared with cohabitation.

45

While these studies address important issues, poverty is one economic outcome
that has not received attention. In this essay, a model similar to those used in the above
studies will be used to assess the EITC’s effects on poverty status among single women.
The impact on poverty is fundamental to the success or failure of the EITC, especially in
light of welfare reform. The only previous investigation of the effects of the EITC on
poverty status is Neumark and Wascher (2001). The Neumark and Wascher study
focused on transitions out of poverty following EITC using matched CPS records. This
methodology allowed them to estimate the impact of EITC expansions, at both the state
and federal level, on labor supply and poverty. However, the methodology can only
capture the effect that takes place in the year of an EITC expansion. Any delayed
reaction will not be included in the estimated effect of the expansion.

This study will use pooled cross-sectional data to trace the impact of the EITC
over time while allowing for a longer adjustment period. There is a scarcity of empirical
work on the impact of the EITC on poverty. The results in Chapter 4 are intended to help
address this void in the literature.

This study will use a model based on the differences-in-differences framework to
investigate the impact of the EITC on the probability that single mothers work or are in
poverty. This study relies on year-to-year variation in the federal and state earned
income credits to identify any impact among a sample drawn from the March CPS. The
model is shown to produce estimates consistent with the existing literature on labor-
market participation. It is important to evaluate whether the increase in work

participation results in reduced poverty.

46

The ﬁrst set of regressions examines the effects of various policies, including the
EITC, on the labor-market participation of single mothers. The second set of regressions
examines the effects of the EITC on the poverty status of single mothers, looking at both
100 percent and 150 percent of poverty. The results for poverty status are presented in

Chapter 4.

3.2. IDENTIFICATION

Studies of the earned income tax credit have taken a number of approaches.28
Browning (1995) and Dickert et al. (1994) used micro simulations, attempting to look at
the structural implications in a sort of general-equilibrium format. Meyer and
Rosenbaum (2001) have focused more on a model that is partial equilibrium, focusing on
the prices and wages offered in the labor market. The study is rigorous in its inclusion
and speciﬁcation of the various tax, welfare, and health-care policies affecting labor
supply.

Eissa and Liebman chose the natural-experiment method, relying on the changes
in environment faced by the subject group compared with a control group. This method
is less complete in the detailed structure of the model, and relies on the assumption the
policy change is the only factor (or at least the predominant factor) driving any change in
behavior. One potential source of bias from using this approach would occur if, over
time, the probability of the measured outcome (in this study labor-force participation or
poverty) becomes greater due to either economic or policy changes for which there are no

explicit controls. For example, if having children restricts labor-force participation more

 

2’ Hotz and Scholz, 2000.

47

or less as time goes on, the estimated effect on participation of the policy change will be
biased.

A difﬁculty that occurs in these difference-in-differences studies is that there may
be inﬂuences affecting the measured outcome that are omitted from the study, either
through oversight or due to an inability to quantify the desired inﬂuence. For example, in
studies examining the effect of the earned income tax credit on the labor-market
participation of single mothers, any inﬂuence of the minimum-wage rate has generally
been omitted, with the exception of the study by Neumark and Wascher (2001).

Recent policy has emphasized getting single mothers to work in order to receive
public assistance, both at the state and federal levels. Both the popularity of welfare
waivers during the 1990s and the push toward welfare reform reﬂect this. It is difﬁcult to
quantify adequately this increased emphasis on work. This study includes controls for
minimum wages and the aggressiveness of welfare reform, in an attempt to present a
more complete policy framework.

In a natural-experiment framework, any omitted inﬂuence due to the increased
emphasis on work will likely lead to an upward bias on the estimated effect of other
variables. One potential immeasurable source of bias is the impact of the public
discussion of welfare reform. It would seem unreasonable to believe that public
discussions of revising the welfare state to encourage work in order to receive any
assistance would not encourage some recipients to exert extra effort to search for work.
However, the quantiﬁcation of this inﬂuence is impossible.

In addition, some states were aggressively pursuing reforms, thanks in part to

federal welfare waivers. Wisconsin spent much of the decade of the 199OS attempting to

48

reform welfare, and likely pursued reform at a level over and above that of most other
states (see Ellwood 2000). The state was a leader in designing alternatives to traditional
public assistance, was among the ﬁrst to require work in order to receive beneﬁts, and
has the most generous state earned income credit for large families, those with 3 or more
children. These policies unique to Wisconsin should raise labor-market participation, and
it is likely the natural-experiment framework will attribute more success to the coincident
policies that are under examination, even though the unobservable initiative was
responsible for some of the effect.

In addition, the simultaneity of the changes in multiple programs in the mid-to—
late 1990s has made it difﬁcult to separate out ﬁnal effects. Since 1993, the minimum
wage has increased, the eamed income tax credit has almost doubled, states began
widespread experimentation with revisions to the AFDC program, and economic growth
greatly accelerated, especially after 1995. Given their potential interactions, it is difﬁcult
to specify completely a model to control for these simultaneous changes, and thus to
identify and estimate their effects.

A strong job market expands opportunities for all potential workers as the demand
for workers increases, thus improving the probability of success for almost all job
applicants. Higher minimum wages raise the return to work. This will encourage more
low-wage workers to look for work at the higher minimum wage. Now combine the
above factors with an expansion of the EITC and welfare policy shifts toward
encouraging work. It may be difﬁcult to obtain precise estimates of their individual
effects if changes occur in these policies coincidentally. The identiﬁcation of speciﬁc

effects is more precise if there is variation among states in some of these factors, such as

49

the addition of state EITCs. State EITCs provide two key sources of identiﬁcation. First,
states that add EITCs present a source of variation from those states that do not add an
EITC. Second, because state EITCs usually are implemented as a percentage of the
federal credit, an expansion in the federal credit will increase the overall EIT C by more in
a state that has a separate state EITC. The estimation of the impact of nearly coincident
policy changes is also aided if certain policies or economic conditions affect certain
groups differently, such as the EITC affecting families with children and the minimum
wage affecting all low-wage workers.

Welfare reform may eventually cause the measured response to the expansions of
the EITC from the low-income population to be greater than it otherwise would be. As
time limits for welfare beneﬁts are imposed, individuals who may not have otherwise
entered the labor market may now do so, for reasons separate from the EITC. Near-
simultaneous modiﬁcations of programs targeting similar groups should result in larger
numbers of new labor-market participants than would be attributable to the EITC alone.
This will pose challenges for studies evaluating time periods encompassing welfare
reform.

Schoeni and Blank (1999) discuss this identiﬁcation problem while attempting to
estimate the effects of welfare waivers and reform. They recognize that attributing any
increase in labor-force participation to welfare reform assumes the full impact of the
1993 expansion of the earned income credit was incorporated in behavior prior to welfare
reform. To allow clearer identiﬁcation of the impact of the EITC, the period immediately
following federal welfare reform (1997-1999) is excluded from this study. It is intended

that this will allow for more precise identiﬁcation of the effects of the EITC.

50

For this study, a policy framework is developed that controls for all of the major
factors believed to be affecting labor-force participation among the low-income
population. There are three sources that allow for speciﬁcally identifying the effects of
the EITC. The ﬁrst source of identiﬁcation is that the EITC is available only to families
with children. This was completely true prior to 1994. After 1994, low-income
individuals are eligible for a small credit. For the estimates reported here, that variation
has been ignored, allowing for continued comparisons between women with children and
those women without children. The second source of identiﬁcation is the variation in the
federal EITC over time. The EITC was relatively small (both absolutely and at the
margin) during the mid-19805. The credit was increased three times between 1986 and
1993, making the EITC the largest federal expenditure directed at low-income families.
In addition, the credit has gone from a single credit available to all families to separate
credits aimed at families of different sizes. The variation in the value of the EITC over
the sample period provides a source of identiﬁcation. The ﬁnal source of identiﬁcation is
the difference in the value of the EITC across states. By 1994, seven states had adopted
their own EITCs. The variation in the value of state credits, both across states and over
time, becomes a third source of identiﬁcation.

To test the robustness of the initial estimates, the data sample was restricted to a
couple of subsamples in some of the speciﬁcations that are reported. First, the sample
was restricted to include only women with a high-school education or less. Restricting
the sample to women with no more than a hi gh-school education results in a remaining

sample of 41,708 observations. This group has a greater concentration of women eligible

51

for the credit than the full sample. It would be expected, and prior research has found,29
that the impact of the EITC on labor supply is greater for single mothers with low
educational attaimnent.

Second, to evaluate whether New York’s poor labor market condition at the time
of enacting an EITC adversely affects the estimated effects of state-level EITCs
generally, estimates were obtained excluding the New York observations. The
motivation for these estimates will be discussed below in greater detail. As will be seen
below, there is some evidence that the relative size of New York, when compared to the
other states with EITCs during this sample period, does reduce the estimated

effectiveness of state EITCs at increasing participation and reducing poverty.

3.3. MODEL

The differences-in-differences approach is used here, combined with an attempt to
expand the control variables beyond those used previously, most notably by the addition
of minimum wages and some controls for changes in state welfare policy. This study will
use this approach to assess the impact of the EITC on labor-force participation and
poverty. Increased labor-market participation will, it is desired, decrease the reliance
single mothers place on public assistance, while allowing them over time to remain out of
economic distress or poverty.

The fundamental model that is estimated is the probit equation:

 

29 For example, see Eissa and Liebman (1996) and Meyer and Rosenbaum (2000).

52

P(poit = 1) = P(a + PZit + YYit + boNumber of children + 51(EITC control) +

62(Children>0 x EITC control) + 63(Minimum wage)) + 8,,

where poit is a dummy variable measuring the speciﬁc policy outcome under

examination, either labor-force participation or poverty status. The dummy variable for
labor-market participation is equal to one if a woman reported working one hour or more
during the year. The dummy for poverty status is equal to one if the family’s income is
below the poverty threshold, based on her family’s size. 8 is a random error term that is

assumed to have a standard normal distribution.

Zit is a vector of control variables that includes age and its square, race, unearned

income, total number of children, number of children under 6, and the level of
educational attaimnent. There are also interaction controls for women with low
educational attainment who also have children, nonwhite women with children, and

nonwhites with low educational attainment. Year dummies are also included in Z“.

These variables control for inﬂuences independent of government policy changes that are
not accounted for otherwise and that may inﬂuence the labor-market participation of all
women, including both treatment and control groups.

Y“ is a vector of control variables that reﬂect the labor-market conditions in the
state, including the state-level unemployment rate, the growth rate of state personal
income, state dummies, and, for women with children only, the maximum grant from
AFDC. The minimum-wage variable is the higher of the federal or state minimum wage,

divided by the statewide average wage in manufacturing. Using this ratio provides a

53

measure of minimum wages scaled relative to labor market conditions in individual
states. This is similar to a common measure used in studies of the effects of minimum
wages (see Leighton and Mincer (1981) and Cunningham (1981)). The difference is that
the ratio of minimum to average wages was not multiplied by the ratio of covered
workers. This implicitly incorporates an assumption that the coverage rates for each state
are identical and constant over time. Estimates from the CPS indicate that between 96.5
percent and 98.2 percent of working women earned more than the federal minimum wage
from 1982 to 2000. In addition, the US. Department of Labor (DOL) submits a periodic
report to Congress entitled Minimum Wage and Overtime Hours Under the Fair Labor
Standards Act.30 From 1981 to 1999, the DOL reports a small increase (from 17.4
percent to 18.5 percent) in the percentage of private industry wage and salary workers
who are exempt from the Fair Labor Standards Act, which contains the federal minimum-
wage provisions." The percentage of private wage and salary workers who were not
exempt under the FLSA and not covered by the minimum-wage law declined from 12.0
percent in 1981 to 8.3 percent in 1999. The increase in coverage was larger among all
wage and salary employees. This is due to the inclusion of state and local government
workers under the F LSA. However, this decline is unlikely to have had much empirical
signiﬁcance, since most government employees are paid more than the federal minimum
wage. Given the relatively consistent breadth of the provisions of the FLSA during the
period included in this sample, the coverage ratio was not included in the minimum-wage

variable. For the purposes of this study, the minimum-wage variable is included to

 

’0 Earlier editions of the report were entitled Minimum Wage and Maximum Hours Standards Under the
Fair Labor Standards Act.

3 ' Section 13(a)(1) of the FLSA contains exemptions for executive, administrative, or professional
employees, as well as outside salespeople.

54

control for broad variation across states in the relative magnitude of federal and state
minimum-wage laws. The theoretical examination and empirical quantiﬁcation of
coverage that is crucial to in-depth research on the minimum wage is less relevant to the
goals of this study.

The major welfare program in the US. providing cash assistance prior to 1998
was Aid to Families with Dependent Children (AF DC).3 2 The program was administered
by the states in accordance with federal regulations and requirements. Beginning in the .
early 19605, the Secretary of Health and Human Services could waive some of the
requirements, to allow states to experiment with changes that would further the goals of
the program. At ﬁrst, little use was made of this authority. During the 19805, the Reagan
administration used this authority, but on a limited basis. Most of the waivers affected
small parts of a state’s welfare caseload.

The use of waivers accelerated during the ﬁrst Bush administration, with the ﬁrst
broad statewide waivers being granted in 1992. Statewide waivers were granted in 1992
to New Jersey, Oregon, Michigan, Utah, and Califonria. Over the next four years, 43
new waivers were granted to 33 states. The waivers allowed states to impose limits on
the length of time welfare recipients could receive beneﬁts, impose a deadline for
recipients to obtain a job, lower beneﬁts when families receiving beneﬁts had additional
children, require recipients to participate in training and education programs, and reduce
the high rates of beneﬁt reductions.

There is a limited correlation between welfare waivers and state EITCs. Only
four of the ﬁrst seven states to adopt earned income credits had received any statewide

welfare waiver by 1997. The state with the most aggressive credit, Wisconsin, had

55

received two waivers. Vermont, Maryland, and Iowa were the other states with waivers,
while Minnesota, New York, and Rhode Island had not received a statewide waiver.

In 1996, the Personal Responsibility and Work Opportunity Reconciliation Act
(PRWORA) was enacted. This landmark legislation of welfare reform repealed the Aid
to Families with Dependent Children program. The replacement program, a block-grant
system that provides states with great ﬂexibility in providing beneﬁts, began to be
implemented in the fall of 1996. Because the focus of this paper is on the EITC, the
entire period following the enactment of PRWORA is excluded. Inclusion of this time
period would increase the difﬁculty in identifying the effects of the EITC, by increasing
the number of coincident programs aimed at similar populations.

The speciﬁcation of welfare waivers used here is similar to the one used by
Neumark and Wascher (1998), and uses data taken from the Council of Economic
Advisors (1997). The variable is equal to the number of months a state had a statewide
waiver implemented during the year, divided by 12. This would include earnings
disregards, work requirements, or term limits for beneﬁts. The variable is only nonzero
for women with children, since it would be expected that welfare waivers might affect the
labor-market participation of women with children, but should have no direct effect on
women without dependent children. As mentioned above, welfare waivers had become
relatively widespread by the end of the period.

The estimated effect of the EITC is modeled through two different controls. The
ﬁrst control quantiﬁes state-speciﬁc subsidy rates, incorporating both federal and state-

level EITCs. The subsidy rate equals the wage-subsidy rate for which the woman would

 

’2 lnfonnation on AFDC waivers is taken from Council of Economic Advisors, 1997.

56

be eligible, based on the number of eligible children living with her and her state of
residence. If a woman with three eligible children were living in Michigan in 1985, the
subsidy rate would equal 0.11. A woman in similar circumstances in 1995 had a subsidy
rate of 0.36. A woman with three children living in Wisconsin in 1995 had a subsidy rate
of 0.54 when adding in the 50 percent Wisconsin earned income credit in effect at that
time.

The second speciﬁcation uses a variable that measures the federal wage-subsidy
rate a woman would face upon entering the labor market. This differs from the ﬁrst
speciﬁcation by separating the federal and state wage-subsidy rates. This measure
increases over time as the EITC expands. Beginning in 1991, the federal earned income
credit provided additional beneﬁts to women with multiple eligible children. From 1991
to 1993, the additional credit was very small. The credit rate was slightly higher for
multiple-child families. However, the higher credit was applied to the same amount of
earnings, resulting in a small additional potential credit ($43 in 1991, $60 in 1992, and
$77 in 1993). Since 1994, a higher credit rate and wider subsidy range are available to
multiple-child families, so the potential maximum credit for a family with two or more
children is more than 50 percent larger than the potential credit for families with one
child in 2001. In this speciﬁcation, separate wage-subsidy rates were included to control

for state EITCs.

3.4. DATA

The data used for this investigation are the Annual Demographic and Income

Supplement of the Current Population Survey, completed in March of each year. This

57

ﬁle contains complete detail for each household, as well as annual employment history
and earnings. The data used for this study encompass all of the major expansions of the
federal EITC, covering 1982 through 1996. The initial years of major expansions (1986-
87, 1991, and 1994) were not included. The sample period also includes two increases in
the federal minimum wage.

The sample consists of all single women between the ages of 25 and 50, inclusive.
The total number of records for these years is 83,935. The sample is restricted to women
beyond the normal ages of school attendance, in order to reﬁne the focus on labor-market
participation. The older population used here should be less likely to exhibit behavioral
changes, i.e., a 40-year old woman out of the work force is less likely to enter the work
force than a woman who is 20 years old. There are a couple of potential reasons to
support this generalization. An older woman is more likely to have been unemployed for
a long time, simply because of her age. Longer periods of unemployment would seem to
increase the probability that unemployment would become permanent, because of skill
deterioration. Also, employers may expect greater labor-market experience from an
older applicant than a younger one.

Excluding the younger population also restricts the sample to women more
established in their labor-market patterns. Younger women are more likely to attend
school than older women, regardless of their maternal situation. While it is entirely
possible that an individual over the age of 24 may ﬁnish high school or attend college,
the younger age groups are far more likely to pursue those options. In addition, the
impact of EITCs, welfare policies, and minimum wages on school attendance is the topic

of Chapter 6.

58

Retirement and disability grow in importance as women age. Also, minor
children become less common as women exceed 50 years of age. Including older women
in the sample, when their labor-market behavior is less likely to be affected by the earned
income credit, would potentially obscure the impact of various policies aimed at single
mothers.

Annual poverty estimates were taken from the Historical Poverty Tables available
from the Bureau of the Census. Unemployment rates, population estimates, and the
Consumer Price Index were taken from the Bureau of Labor Statistics data ﬁles, and
personal income was obtained from The Survey of Current Business (U .S. Department of
Commerce, (2000)).

State-level minimum wage data were obtained from the annual labor legislation
update from The Monthly Labor Review. Annual data on maximum AFDC beneﬁts for
each state came from The Green Book (various years) and Aid to Families with
Dependent Children (U .S. Department of Health and Human Services, (various years)).
Information on welfare waivers was obtained from Explaining the Decline in Welfare
Receipt, 1993-1996, a technical report published by the Council of Economic Advisers.

Some general statistics on the sample are presented in Table 3.1. Table 3.2
presents a summary of the dependent variables. It is straightforward to see that women
with children (thus eligible for the EITC) experienced larger increases in participation

and decreases in poverty, especially after 1993.

59

3.5. ANALYSIS OF THE RESULTS

The ﬁrll results are presented in Tables A-6 through A-17. The results are
reported as partial derivatives evaluated at the sample means. Each speciﬁcation is
estimated once for each of the three different EITC variables. Each appendix table
presents the estimates for one speciﬁcation and one set of EITC controls for each of the
three dependent variables: labor-force participation, poverty, and near-poverty status.

The ﬁrst speciﬁcation is the basic model.

P(poit = 1) = P(oc + pZit + YYit + boNumber of children + 51(Children>0 x

EITC control) + 62(Children>0 x Welfare waivers) + 63(Minimum wage))
+ Sit
The vectors lit and Yit are described above. 8 is once again a random error term

that is assumed to have a standard normal distribution. There are three speciﬁcations of
the EITC variable. The ﬁrst is the combined state and federal EITC subsidy rate faced by
a woman with children upon entering the labor market. The second EITC control
separates the state and federal subsidy rates into two variables. The third set of EITC
variables has separate variables for the federal and refundable state EITCs.

The next speciﬁcation re-estimates the above model using the subsarnple of
women with no more than a high school education. Year and child interactions were then

added to the basic model. The revised model becomes:

60

P(poit = 1) = P(a + BZit + YYit + 60Number of children + 51(Children>0 x

EITC control) + 52(Children>0 x Welfare waivers) + 53(Minimum wage)

+ 54(Year dummy x Children>0)) + 8,,

where 8 is once again a random error term that is assumed to have a standard
normal distribution. Finally, estimates were obtained for the model including the year
and child interactions using the low-education subsample. A number of variations are
then estimated to test the robustness of the estimates obtained using the basic model. The
results are reported for the EITC, and occasionally welfare waivers, minimum wages, and

welfare reform. However, the full results are omitted.

3.5.1. Estimation Results for General Controls

The results for the basic model are presented in the ﬁrst set of columns of Tables
A-6 through A-8. The estimates are generally consistent with the standard theory from
labor economics and are also consistent across the different speciﬁcations and with
different EITC controls.

Labor-force participation increases with age, but at a declining rate. Nonwhites
are approximately 6.9 percentage points less likely to participate in the labor market than
whites. The presence of an additional child under 6 years old in the household reduces
labor-market participation by about 5 percentage points.

Education level is a strong predictor of labor-force participation. Women who did
not graduate from high school participate at much lower rates than women with higher

educational attainment. The average participation rate across years for the entire sample

61

is 87.0 percent, while the average for the sample that includes only women completing a
high-school diploma or less formal education is 79.8 percent. Women with less than a
diploma, a subset of the high-school-or-less sample, participate at a signiﬁcantly lower
rate. Overall, only 61.9 percent of women with less than a high-school diploma
participate in the labor market.

The estimated effect of not completing high school is a reduction of participation
of approximately 15.3 percentage points, compared with high-school graduates.
Completing at least some college results in higher participation, with single women with
some college attendance estimated to be approximately 4 percentage points more likely to
participate. College graduates are estimated to be 4 percentage points more likely to
participate in the labor market than those with some college. This implies that the
probability of labor-force participation for women who have completed at least a
bachelor’s degree is eight percentage points higher, all else equal, than for women with
only a high-school education, and more than 23 percentage points higher than for women
without a high-school diploma.

Higher amounts of unearned income (such as interest or transfer payments) lower
participation among unmarried women, other factors unchanged. An increase of $1,000
in real non-labor income decreases the probability of labor-market participation by an
average of 0.3 percentage points.

Higher levels of public assistance within each state are associated with reduced
participation. The coefﬁcient is consistently negative in all speciﬁcations and signiﬁcant

at standard conﬁdence levels. However, the coefﬁcient is relatively small, with a $100

62

increase in real monthly beneﬁts producing less than a 2-percentage point decline in
participation.

An increase of one percentage point in the unemployment rate is estimated to
reduce the probability of employment by 1.1 percentage points, other factors unchanged.
Consistent with the ﬁndings of Meyer and Rosenbaum (2001), the effect of higher
unemployment is not especially burdensome on single mothers. The estimates indicate
that higher unemployment is not associated with any signiﬁcantly larger effect on women
with children.

Higher levels of personal income growth are associated with lower levels of
labor-force participation. Personal income growth is measured as the annual percentage
change in personal income in each state in the reference year covered by the CPS. One
possible reason for this is that personal income growth may be high coming out of a
recession, when many people are not employed.

As a rule, the estimated coefﬁcients for the demographic and economic controls
are consistent across the different speciﬁcations. There are some exceptions for the low-
education subsamples (Tables A-9 through A-l 1). However, in order to focus on the
EIT C, discussion of the estimates for the general demographic and economic controls

will be omitted.

3.5.2. The Impact of the EITC on Participation
Basic Model
Using the basic model, the federal EITC is estimated to increase participation by

between 3.5 and 4.9 percentage points in the period after OBRA 1993, depending on the

63

EITC control variable included in the model. These estimates are signiﬁcant for the
combined federal/state subsidy rate and the separate federal subsidy rate. A summary of
the estimated impact of the EITC on participation using the basic model is presented in
panel A of Table 3.3. The highest coefﬁcient estimates are from the regressions using
only the federal wage-subsidy rate, with the estimated partial derivative of0. 122
implying an increase in labor-force participation of 4.9 percentage points among women
with 2 or more children in 1996, all other factors unchanged.

The estimated coefﬁcients imply an increase in labor-market participation of 1.7
percentage points for the TRA-1986 expansion. This result is smaller that the estimates
obtained by Eissa and Liebman, who found an estimated effect of between 1.9 and 2.9
percentage points. This may be due to the older sample used for this sample, given the
strong effect the EITC has on the participation among women under 25, as reported in
Chapter 6. Eissa and Liebman’s sample was unmarried women ages 16 to 44, while the
sample used here is ages 25 to 50. Estimates on this data set using a model similar to the
one used by Eissa and Liebman, and covering a similar time period, produced smaller
effects than the results they reported.33 The estimated effect of the EITC on participation
was positive. However, the increase in participation was smaller than the results reported
by Eissa and Liebman. For women with no more than a high—school education, the
estimated effect of the EITC expansion in 1986 was a 2.2-percentage—point increase,
close to the 2.6-percentage-point increase reported by Eissa and Liebman for women with

only a high-school education.

 

33 The sample period used by Eissa and Liebman was 1984-1986 and 1988-1990. The data used here are
from 1982-1985 and 1988-1990.

64

The estimates for the effect of state ElTCs on labor-market participation are
negative (-0.l37) and insigniﬁcant for the basic model. Similar qualitative results (-
0.049) were obtained using only refundable state credits. Since state-level credits
unambiguously increase the incentives to participate in the labor market, these results are
surprising. These results are also contrary to those of Neumark and Wascher (2001). For
families without a worker in the ﬁrst year, Neumark and Wascher found that state-level
credits increase signiﬁcantly the probability there was a worker in year two.

The methodology used here is quite different from that employed by Neumark
and Wascher. In this study, control variables are included for the obvious factors that are
believed to affect labor supply, including the EITC. The key source of identiﬁcation is
the discrete policy shift that took place each time the federal EITC was increased.
However, any omitted inﬂuence that is correlated with a variable included in the model
will result in biased estimates. Neumark and Wascher attempt to eliminate this potential
bias by including an interaction term for women with children by each year of the
sample. These year/child interaction terms capture any general inﬂuence affecting
women with children. In the Neumark and Wascher framework, the identiﬁcation
strategy relies on variation between year one and year two in the federal and state EITCs.
However, the year/child interaction variables are going to have a high degree of
collinearity with the level of the federal EITC, which is used here. In fact, the year/child
interactions are the identiﬁcation strategy used by Eissa and Liebman to identify the
effects of the EITC. Including year/child interactions in this study may reduce the

explanatory power of the federal EITC variables.

65

However, to test whether the estimated effect of state-level EITCs is better
identiﬁed when year/child interactions are included, the model was re-estimated using
these variables. The results are presented in panel B of Table 3.3. The estimate for the
combined federal/state subsidy rate becomes smaller (from 0.102 to 0.068) and
insigniﬁcant. The reason for the insigniﬁcant estimate is a large increase in the standard
error of the estimated coefﬁcient. When the federal and state subsidy rates are estimated
separately, a large increase in the estimated coefﬁcient for the federal EITC more than
offsets an increase in the standard error, so the estimated coefﬁcient remains signiﬁcantly
different from zero at all traditional signiﬁcance levels. The estimated coefﬁcient for the
federal EITC is almost three times as large as the estimates reported in panel A.

Contrary to the estimates for the federal EITC, the estimates for state EITCs
presented in panel B of Table 3.3 are virtually unchanged from those in panel A. The
consistency of the estimated coefﬁcients and standard errors for state EITCs provide
additional evidence that state EITCs are not associated with increased labor-force
participation. The results from estimating the base speciﬁcation that are presented in
panel A are conﬁrmed by those in panel B.

The increased participation of single mothers in the labor market is an important
indicator of the success of the EITC. Another measure of the impact of the EITC is how
the EITC affects the earnings of single mothers, compared with single women without
children. A traditional measure of the elasticity of eanrings with respect to the EITC
would be difﬁcult to obtain, since the EITC has a signiﬁcant effect on the participation
decision, and those women entering the labor force will generally earn less than women

already in the labor force. Thus estimates of this elasticity conditional on participation in

66

the labor force are negative. This is consistent with the EITC encouraging labor-force
participation among low-wage women with children.

As an alternative measure, the change in average earnings for women with
children can be compared with the change in average earnings for women without
children. The estimates are presented in Table 3.4. For women in the sample with no
more than a high school education, the percentage change in real earnings for women
without children was subtracted from the percentage change in real earnings for women
with children for each EITC increase (1986, 1990, and 1993). The difference in earnings
growth is then divided by the percentage increase in the EIT C wage-subsidy rate for a
woman with one child. The result represents the percentage increase in the real earnings
of women with children, adjusted for any increase accruing to all women, divided by the
percentage increase in the EITC.

The results indicate that average real earnings for women with children have
increased by a larger percentage than the average earnings of women without children in
each period following an increase in the EITC. The elasticity estimates for the periods
following the 1986 and 1993 increases in the EITC are around 0.1, indicating that a 10
percent increase in the EITC would raise the average real earnings of single mothers by 1
percent over the earnings growth of single women without children. The elasticity
estimate for the 1990 EITC increase is 0.026. This estimate is most likely lower because

of the impact of the 1991 recession.

67

High School or Less Education

The estimated effect of the federal EITC increases when the sample is limited to
women with no more than a high-school education. These estimates are presented in
panel C of Table 3.3. The expansion of the EITC in 1993 is estimated to have increased
participation by from 7.6 to 9.8 percentage points, depending on the speciﬁcation. This is
approximately twice as large as the estimate discussed above for the hill sample, where
the estimates ranged from 3.5 to 4.9 percentage points. The results for this subgroup
reinforce the claim that the EITC is responsible for a signiﬁcant part of the increase in
labor-market participation among single mothers. The sharp increase in the size of the
estimated coefﬁcients is consistent with the results of Eissa and Liebman, as well as
Meyer and Rosenbaum (2000). Both studies reported signiﬁcant increases in the
estimated impact of the EITC in subsamples with lower educational attainment.

Regressions were also run on the women in the sample with higher educational
attainment than a high-school diploma. These estimates are not reported here, but the
estimates indicate a general correlation between lower participation and the EITC among
women with children and more than a high-school education, the opposite effect from
women with no more than a high-school diploma. This implies that more than 100
percent of the effect observed in the entire sample is due to the impact on women with
relatively low educational attainment. The estimates for the mothers with more than a
hi gh-school diploma are not signiﬁcant at traditional levels.

Among women with lower levels of educational attainment, all state EITCs
continue to have negative coefﬁcient estimates, although the estimates are less precisely

measured (have higher standard errors) and statistically insigniﬁcant. The estimated

68

coefﬁcient for refundable state EITCs does become positive, although the coefﬁcient is
not statistically signiﬁcant.

To further investigate the identiﬁcation of any impact attributable to state EITCs,
year/child interactions were included in this speciﬁcation using the subsample of only
women with a high school education or less. The estimates are presented in panel D of
Table 3.3. The results are very similar to those presented in panel B. The estimated
coefﬁcient on the combined subsidy rate is not signiﬁcantly different ﬁ'om zero. The
coefﬁcient estimates for the separate federal EITC are very close to those presented in
panel B, although the standard errors are larger. Consequently, the estimated coefﬁcients
are not signiﬁcant at the 10—percent conﬁdence level. The collinear nature of the federal
EIT C and the year/child interactions may be more of a problem given the smaller sample
size. The estimated effect of state EITCs is not signiﬁcant for either all state EITCs, or
for just refundable EITCs. However, the sign of the estimated coefﬁcient for refundable

EITCs is positive, as it was in panel C.

New York Excluded

Questions about the estimated effects of the state EITCs led to a new estimation
speciﬁcation. New York is the largest state to have adopted a credit and, as was
mentioned earlier, was facing a much different labor-market situation than were the other
states with EITCs. Some comparisons between New York and the national averages for
unemployment, poverty, and labor-force participation are presented in Table 3.5. Prior to
1991, New York was below the national average in unemployment for every year since

1981. However, 1991 seems to have marked a deterioration in the labor market in New

69

York, as unemployment became more prevalent in New York than in the US. as a whole.
Labor-force participation peaked in 1990 in New York at 62.9 percent, but began to
decline throughout the early 19905. Measured poverty also became more prevalent in
New York during the early 19905, both absolutely and relative to the US. average. As
the discussion in Chapter 2 highlighted, most of the other states with EITCs adopted an
EITC while the state’s labor market was performing better than the US. averages. New
York is clearly unique in that the state labor market was not performing as well as the
overall US. economy at the time the New York EITC was enacted. Given New York’s
relative size compared with the other states with EITCs, it is possible that New York
exerts a negative inﬂuence on the estimated coefﬁcients for state EITCs.

In order to investigate whether New York’s economy was responsible for the
negative correlation between state EITCs and labor supply, the observations for New
York were excluded. By excluding New York, any negative effect on the estimated
coefﬁcient for state EITCs resulting from comparisons with New York was removed. In
addition, to better assess the impact of state EITCs, the model was ﬁrst estimated using
all state EITCs, and then estimated using only refundable state EITCs (excluding New
York in both cases).

The estimates are summarized in panel B of Table 3.3. The estimated effects of
the federal earned income credit increased slightly from those presented in panel A. The
estimated effect of all state EITCs on labor-force participation, all other factors remaining
unchanged, remains negative. The estimate is somewhat smaller in magnitude and
remains insigniﬁcant. However, the estimated effect of reﬁrndable state-level credits

became positive, indicating a positive correlation between refundable state credits and

70

labor-force participation. Although now positive, the coefﬁcient estimates for state
credits were not signiﬁcantly different from zero at traditional levels.

The large shift in the estimates for refundable state EITCs with the New York
observations excluded is consistent with the hypothesis that the New York EITC was
negatively affecting the estimate. The results with New York excluded are more closely
in line with the ﬁndings of Meyer and Rosenbaum (2000) and Neumark and Wascher
(2001) that state EITCs are at least somewhat effective at increasing labor-force

participation among single mothers.

Welfare Reform

It may be true that welfare policy in general leading up to the passage of the
PRWORA led to increased participation. The increased use of waivers may have been an
indicator or reﬂector of modiﬁcations in state policy toward welfare. While the presence
of a waiver would indicate an explicit policy shift, more subtle efforts to reduce the
welfare population, or at least to increase work among recipients, may also have been at
work. If states were attempting to reduce welfare dependency using less explicit
methods, estimates of the effectiveness of EITCs and welfare waivers may be overstated.

In order to control for this, a measure of the emphasis states were placing on
reducing caseloads and encouraging working is necessary, beyond controls for waivers
and the ﬁxed effects captured by state dummies. Ellwood (2000) estimated a model of
welfare participation for each state from 1984 through 1992. Then, using the parameter
estimates from the welfare-participation model and the demographic characteristics of

women in a sample drawn for the March CPS for 1997 and 1998, Ellwood predicts

71

welfare-participation rates for each state for 1997 and 1998. Ellwood then subtracts the
predicted rate from the actual participation rate in the CPS sample for 1997 and 1998.
The resulting difference provides a measure of the aggressiveness of each state in
reforming welfare between 1992 and 1998.34 In order to avoid assuming this difference
between the predicted and actual percentage of single mothers receiving welfare beneﬁts
represents a precise measure of aggressiveness, Ellwood uses the difference to classify
states into three categories. States with an unpredicted decline in welfare participation of
4 percentage points or less were given a value of 1, those states with an unpredicted
decline of between 4 and 10 percentage points were given a value of 2, and those states
with an unpredicted decline of 10 or more percentage points were given a value of 3. For
this study, the measure of aggressiveness for each state is then used as an explanatory
variable to control for differences in welfare policy between states during the mid-19905.
As might be expected, the estimated effect of the federal EITC decreases slightly
with the addition of the welfare reform variable, although all the estimates remain
signiﬁcant. The estimate for the combined subsidy rate decreases from 0.102 to 0.100
when the welfare reform variable is included. This implies an estimated increase in
labor-force participation of 2.1 percentage points for women with two children after the
OBRA-1993 expansion, with the EITC’s overall inﬂuence amounting to approximately a
4.0-percentage point increase in participation among women with two or more children.
A similar decrease occurs when the federal and state EITCs are separated. The estimated
coefﬁcient for the federal EITC decreases from 0.122 without welfare reform included

(panel A) to 0.120 with welfare reform included (panel F). The estimated coefﬁcient for

 

3‘ More corrrpletc explanations of the estimation procedure and the state-by-state estimates are in
Ellwood’s paper and are not repeated here.

72

the federal EITC also declines slightly when only refundable state EITCs are included in
the regression. The estimated effects of state EITCs are unchanged from the base model

when the welfare reform variable is included.

Trends for Women with One Child or Two or More Children

Two trend variables are added to the basic model. These trend variables are
created by interacting a time trend with a couple of dummy variables that indicate the
number of children each woman in the sample has. The ﬁrst trend is for women with one
child. The second trend variable is for women with two or more children. These
variables are included to capture any inﬂuence that is affecting these two groups of single
mothers over time. One advantage these variables have over year/child interactions is
that trends are not as collinear with the federal EITC as year/child interactions. One
disadvantage is that a trend variable will only capture those inﬂuences that are monotonic
over time.

These estimates are presented in panel G of Table 3.3. The inclusion of the trend
variables results in virtually no change in the EITC estimates, whether the federal and
state wage-subsidy rates are combined or separate. The estimates for the federal EITC
remain signiﬁcant at all traditional conﬁdence levels. The estimates for state EITCs are
consistent with those discussed for the basic model, for both all state EITCs and only the
refundable credits. To the extent that the trend variables for one and two-or-more
children capture any inﬂuences that were affecting single mothers during the sample
period, these estimates provide additional evidence that expansions of the BIT C have

resulted in increased participation among single mothers.

73

Logged E] T C Eﬂects

It was mentioned above that the study by Neumark and Wascher (2001) might
understate the ﬁnal effect of the EITC because of the limited response time allowed by
using matched CPS records. If at least some individuals who would be eligible for the
EIT C take time to learn about legislative changes that expand the credit, estimates based
on the one-year response will understate the overall effect. In order to evaluate this
hypothesis, 3 series of regressions were estimated using a two-year lag structure for the
EIT C variables. For one set of estimates, the concurrent EITC subsidy rate was included.
For the second set, the concurrent EITC subsidy rate was excluded.

The results, presented in Table 3.6, provide evidence that the federal EITC has
lagged effects on labor-force participation. In panel A, the estimated coefﬁcient for the
two-year lag is signiﬁcant at ﬁve percent for the combined federal/state subsidy rate and
for the separate federal rate. The estimated coefﬁcients are quite large at approximately
0.25. The estimated coefﬁcients for the current rate are consistently positive but
insigniﬁcant. The estimated coefﬁcients are not small in magnitude, but the standard
errors of the estimates are relatively large. In general, the estimates for the one-year lag
on the EITC are negative and insigniﬁcant. These estimates imply that, for example, the
increase in the federal EITC subsidy rate that occurred in 1994 did not have a signiﬁcant
effect on participation until 1996. The estimates obtained when the concurrent EITC is
excluded also have a similar pattern, as shown in panel B of Table 3.6.

State EITCs also seem to be more effective when lagged effects are considered.
The estimates presented in Table 3.6 indicate that the concurrent effect of state EITCs is

negative, but the lagged effects are positive. For refundable state EITCs, the one-year

74

lagged effect is positive and signiﬁcant. When the current state and federal EITC rates
are excluded, the one-year lagged effect is estimated to be negative, while the two-year
lagged effect is positive.

The use of lagged EITC variables indicates that a signiﬁcant part of the response
to changes in the federal EITC occur two years after the change. In every speciﬁcation
presented in Table 3.6, the estimated coefﬁcient for the federal EITC is positive and
signiﬁcant at 10 percent. The pattern for state EITCs is less clear, but there does appear

to be some positive effect on participation over time.

Comparing the El T C with the Minimum Wage and Welfare Reform

The estimated effect of minimum wages on labor-market participation is negative
consistently, although the estimates are marginally signiﬁcant at traditional signiﬁcance
levels. The estimates are summarized in Table 3.7 and presented in detail in Tables A-6
through A-17. While providing limited evidence pointing to negative employment
effects ﬁ'om higher minimum wages, this evidence is consistent with the textbook model.
The estimates for the base model imply that a 10-percentage—point increase in the
minimum wage relative to the average manufacturing wage would reduce the
employment of single women by approximately 1.2 percentage points.

The evidence on the effectiveness of welfare waivers is mixed. The estimated
effects are summarized in Table 3.7 as well. The estimated coefﬁcients are generally
greater than zero. However, the estimates are small, implying less than a one-percentage-
point increase in employment for a state with a waiver in place for an entire year. The

estimates are statistically insigniﬁcant for each speciﬁcation. There are two explanations

75

for these estimates. One is that waivers are ineffective at encouraging labor supply. An
alternative explanation is that waivers may differ across states by type of waiver, and by
the energy with which the state implements the new policy. The estimates presented here
fail to provide any evidence that welfare waivers, by themselves, increase labor-force
participation.

This evidence is at odds with the evidence presented by Schoeni and Blank
(2000). Schoeni and Blank ﬁnd that waivers signiﬁcantly increased employment among
female high-school dropouts. However, the study does not include controls for the EITC.
Since a large expansion of the federal EITC is coincident with welfare waivers becoming
more widespread, it is possible that some of the employment effects associated with the
EITC’s expansion are being captured by waivers in the Schoeni and Blank estimates.

To test the hypothesis that the omission of an EITC control may account for some
of the difference between the results reported by Schoeni and Blank, and the results
presented here, estimates were obtained with the EITC variables excluded. Table 3.8
shows the estimated coefﬁcients for welfare waivers and the minimum wage when the
basic model from this study is used, but the EITC controls are excluded. As shown in
panel A of Table 3.8, the presence of a welfare waiver for an entire year, other factors
unchanged, increases labor-force participation among single mothers by 2.4 percentage
points. Once again, this estimate is obtained when the EITC variables are excluded. This
provides evidence that the difference in the estimated effect of welfare waivers between
this study and the effect reported by Schoeni and Blank may be due, at least in part, to the
omission of an EITC variable from their study.

The estimates reported by Meyer and Rosenbaum (2001) come from a more

76

complete model than the one employed by Schoeni and Blank. The model used by
Meyer and Rosenbaum produces estimates that indicate that state welfare waivers are
associated with signiﬁcant increases in labor supply, although the effects are modest
when compared with those of the EITC. The differences between the estimates presented
here and those provided by Meyer and Rosenbaum may be due to variations in the
speciﬁcation of welfare policy changes. Meyer and Rosenbaum use two particular
aspects of waivers (time limits for beneﬁts and beneﬁt terminations). These variables
may speak more closely to policies directed at moving welfare recipients into jobs. The
policy variable used here, the presence of any statewide waiver, is more general. A
waiver may have included other programs, such as an earnings disregard, which may
encourage additional labor supply but has no punitive action associated with labor-market
inactivity.

These differences are the exact reason the estimates using Ellwood’s measure of
the aggressiveness with which a state pursued welfare reform were used in the previous
section. That a state received a waiver is not a crucial issue. More important is whether
the state attempted to use the waiver to aggressively induce low-income families to work.
The results obtained using this control variable seem to provide additional evidence that,

at least for this sample, the EITC is a more powerful tool for increasing labor supply.

3.6. CONCLUSION
The results presented thus far are broadly consistent with the existing literature on
the EITC. The EITC is associated with increased labor-market participation among

single mothers, accounting for an increase of between 3.5 and 9.8 percentage points in

77

the participation rate of single mothers. There is some evidence that the EITC’s impact
occurs with a lag. While the EITC is positively correlated with labor supply, other
policies aimed at improving the welfare of low-income families do not have a clear
correlation. There is no evidence that higher minimum wages are positively correlated
with increased participation among single mothers. The evidence here suggests that a 10-
percentage-point increase in the relative minimum wage lowers participation (deﬁned as
working in the past year) by one percentage point. The evidence presented here for

welfare waivers points to small and insigniﬁcant effects.

78

Table 3.1

Sample Characteristics

Variable

Age

Nonwhite
Kids

Child < 6
Degree

Some College
Diploma

Less than High School
Hours > 0
Below Poverty

Minimum Wage as Percent
of Hourly Earnings

Real AFDC Beneﬁt

Number of Observations

79

Average

35.1

24.5%

42.1%

17.1%

25.8%

51.3%

35.7%

13.1%

87.0%

20.4%

35.8%

$479.32

83,942

Standard
Enter

0.0291

0.0018

0.0019

0.0019

0.0017

0.0020

0.0019

0.0013

0.0013

0.0016

0.0002

0.7670

1982
1983
1984
1985
1988
1989
1990
1992
1993
1995
1996

1982 - 85

1988 - 93

1995 - 96

Table 3.2

Labor-Force Participation

Entire

Sample

84.4%
84.8%

‘ 85.4%

86.3%
87.8%
87.8%
87.9%
86.3%
86.4%
89.0%
88.7%

With

Children

75.3%
75.0%
77.1%
77.5%
78.3%
79.7%
79.9%
77.4%
77.8%
82.9%
83.1%

76.3%

78.6%

83.0%

80

Participation and Poverty Statistics Over Time

Poverty
Entire With
Sample Children
23.4% 37.8%
23.0% 38.1%
22.6% 37.1%
21.9% 38.1%
18.8% 32.2%
18.5% 31.2%
18.7% 32.0%
20.8% 35.2%
21.3% 34.8%
18.0% 29.6%
19.0% 31.1%
37.8%
33.1%
30.4%

Table 3.3

Estimated Effects of the EITC on Labor Supply

Combined
A. Basic Model
Combined Subsidy Rate 0.1018 **
(0.0223)
Separate Federal/State Rates
Refundable State EITC Rate
B. High School or Less
Combined Subsidy Rate 0.2239 **
(0.0440)

Separate Federal/State Rates

Reﬁmdable State EITC Rate

C. Including Year/Child Interactions

Combined Subsidy Rate 0.0678

(0.0897)

Separate F ederal/State Rates

Refundable State EITC Rate

Federal

0.1222 **
(0.0244)

0.1174 ..
(0.0244)

0.2495 **
(0.0483)

0.2452 **
(0.0483)

0.3409 **
(0.1348)

0.3374 .4
(0.1348)

D. Including Year/Child Interactions - Low Education Sample

0.1014
(0.1636)

Combined Subsidy Rate

Separate Federal/State Rates

Refundable State EITC Rate

E. Excluding New York Observations

Combined Subsidy Rate 0.1163 **

(0.0238)
Separate Federal/State Rates

Refundable State EITC Rate

81

0.3276
(0.2558)

0.3258
(0.2557)

0.1301 "
(0.0250)

0.1258
0.0248

All State
EITCs

-01373
(0.1143)

-0.0673
(0.2127)

0.1340
(0.1148)

-0.0561

(0.2133)

-0.093 1

(0.1193)

Refundable
EITCs

-00490
(0.1710)

0.0469
(0.3073)

-0.0367
(0.1730)

0.0748
(0.3097)

0.1954
(0.2268)

Table 3.3 (cont'd).

F. Includes Welfare Reform

Combined Subsidy Rate 0.0996 **
(0.0223)
Separate Federal/State Rates 0.1196 **
(0.0245)
Refundable State EITC Rate 0.1147 **
(0.0245)
G. Basic Model using Trends for 1 and 2 or More Children
Combined Subsidy Rate 0.0963 **
(0.0267)
Separate Federal/State Rates 0.1230 **
(0.0295)
Refundable State EITC Rate 0.1175 *"
(0.0296)

** Indicates signiﬁcant at 5% conﬁdence level.
* Indicates signiﬁcant at 10% conﬁdence level.

82

-0.1329
(0.1145)

-0.1390
(0.1 144)

-0.0408
(0.1715)

.0.0499
(0.1714)

Table 3.4

Estimated EITC Elasticity of Real Earnings

For Women with No More Than a High School Education

Increase
in Federal
EITC
1986 EITC Increase 27.3%
1990 EITC Increase 32.1%
1993 EITC Increase 83.8%

Increase in
Earnings
for Women
with
Children

9.6%
-1 3.9%

12.7%

83

Increase in
Earnings
for Women
without
MB

6.4%
-l4.7%

4.8%

Difference
in Growth

of Earnings

3.2%
0.8%

7.9%

Estimated
LIN—"m

0.117
0.026

0.094

Table 3.5
Comparison of New York vs. U.S. Averages

 

 

 

 

 

 

Unemployment Poverty Participation
lg; ES, New York LL; New York LS4 New York
1981 7.6% 7.6% 14.0% 14.4% 63.9% 59.8%
1982 9.7% 8.6% 15.0% 14.8% 64.0% 59.4%
1983 9.6% 8.6% 15.2% 15.8% 64.0% 59.4%
1984 7.5% 7.2% 14.4% 16.0% 64.4% 59.3%
1985 7.2% 6.5% 13.6% 15.8% 64.8% 60.7%
1986 7.0% 6.3% 14.0% 13.2% 65.3% 61.2%
1987 6.2% 4.9% 13.4% 14.3% 65.6% 61.7%
1988 5.5% 4.2% 13.0% 13.4% 65.9% 61.8%
1989 5.3% 5.1% 12.8% 12.6% 66.5% 62.9%
1990 5.6% 5.3% 13.5% 14.3% 66.5% 62.9%
1991 6.8% 7.3% 14.2% 15.3% 66.2% 62.2%
1992 7.5% 8.6% 14.8% 15.7% 66.4% 61.7%
1993 6.9% 7.8% 15.1% 16.4% 66.3% 61 .7%
1994 6.1% 6.9% 14.5% 17.0% 66.6% 61.3%
1995 5.6% 6.3% 13.8% 16.5% 66.6% 60.9%
1996 5.4% 6.2% 13.7% 16.7% 66.8% 61.6%
1997 4.9% 6.4% 13.3% 16.5% 67.1% 63.1%
1998 4.5% 5.6% 12.7% 16.7% 67.1% 63.2%
1999 4.2% 5.2% 11.8% 14.1% 67.1% 62.9%

Sources: U.S. Bureau of Labor Statistics, U.S. Department of Labor; and U.S. Census

Bureau, and U.S. Department of Commerce.

84

 

Table 3.6
Estimated Effects of the EITC on Labor Supply
Using Lagged EITC Controls

All State Refundable
Combined Federal EITCs EITCs
A. Including Concurrent and Two Lagged Rates
1. Concurrent 0.1002
(0.1464)
Lagged One Year -0.1722
(0.2256)
Lagged Two Years 0.2493 **
(0.1416)
2. Concurrent 0.1748 -0.5876
(0.1603) (0.5168)
Lagged One Year -0.2350 0.1838
(0.2473) (0.7997)
Lagged Two Years 0.2554 *"' 0.4500
(0.1530) ' (0.5447)
3. Concurrent 0.1755 -1.6382 **
(0.1599) (0.6432)
Lagged One Year -0.2505 1.8233 **
(0.2471) (1.0925)
Lagged Two Years 0.2735 ** 0.1829
(0.1529) (0.7748)
B. Including Two Lagged Rates
1. Lagged One Year -0.0310
(0.0967)
Lagged Two Years 0.2120 *
(0.1320)
2. Lagged One Year 0.0082 -0.5485
(0.1045) (0.4124)
Lagged Two Years 0.1875 * 0.5676
(0.1411) (0.5199)
3. Lagged One Year -0.0057 -0.4650
(0.1046) (0.5910)
Lagged Two Years 0.1967 * 0.7874
(0.1410) (0.7511)

*“ Indicates signiﬁcant at 5% conﬁdence level.
* Indicates signiﬁcant at 10% conﬁdence level.

85

Table 3.7

Estimated Effects of Waivers, Welfare Reform, and the Minimum Wage

A. Using Trends for Families
with Children
Labor-Force Participation

Below Poverty

Below 150 Percent of Poverty

B. Controlling for Aggressiveness
of Welfare Reform

Labor-Force Participation

Below Poverty

Below 150 Percent of Poverty

C. Excluding New York
Labor-Force Participation

Below Poverty

Below 150 Percent of Poverty

Welfare
Waivers

0.0074
(0.0072)

.0.0157
(0.0098)

.0.0035
(0.0126)

0.0092
(0.0072)

.0.0177
(0.0098)

-0.0090
(0.0126)

0.0028
(0.0071)

-0.0135
(0.0099)

0.0032
(0.0127)

" Indicates signiﬁcant at 5% conﬁdence level.
* Indicates signiﬁcant at 10% conﬁdence level.

it

86

Welfare
Reform

A ressiveuess

0.0065 **

(0.0036)

-0.0068
(0.0047)

-0.0130 "

(0.0058)

*

Minimum

Wage

-0.1 199
(0.0724)

0.0355
(0.0992)

0.2230
(0.1212)

-0.1084
(0.0728)

0.0253
(0.0995)

0.2111
(0.1219)

-0. 1086
(0.0703)

0.0333
(0.0992)

0.2220
(0.2169)

**

**

**

Table 3.8
Estimated Effects of Waivers and the Minimum Wage

with no EITC Controls
Welfare Minimum
Waivers Wage
A. Base Model
Labor-Force Participation 0.0237 ** -0.1346 “
(0.0063) (0.0721)
Below Poverty -0.0341 ** 0.0525
(0.0087) (0.0989)
Below 150 Percent of Poverty -0.0239 ** 0.2445 "
(0.0110) (0.1212)
B. Including Year/Child
Interactions
Labor-Force Participation 0.0083 -0.1 176 *
(0.0072) (0.0723)
Below Poverty -0.0182 ** 0.0317
(0.0099) (0.0990)
Below 150 Percent of Poverty -0.0067 0.2253 "
(0.0128) (0.1211)

** Indicates signiﬁcant at 5% conﬁdence level.
* Indicates signiﬁcant at 10% conﬁdence level.

87

CHAPTER 4

EFFECT OF THE EITC ON POVERTY

4.1 INTRODUCTION

The focus here now shifts to the EITC’s impact on poverty. This subject has only
been investigated once in an empirical evaluation (Neumark and Wascher (2001)). The
tests for poverty will be similar to the tests for labor-market participation, and will be
distinct from the approach in the previous work. Neumark and Wascher used matched
CPS records for families. This allowed them to evaluate the impact of expansions in the
EITC over a two-year period. The sample here is not matched, and is composed of all
single women age 25 to 50. This allows for an evaluation of the impact of the EITC over
a longer time period, covering the three EITC expansions between 1986 and 1993. It also
focuses attention on the subgroup of the population who are often cited as the target of
welfare reform.

Two distinct poverty standards will be used. The model attempts to describe the
probability of a woman being below 100 or 150 percent of the poverty threshold. Both
the sample and the estimation techniques remain the same as those discussed above for

labor supply. The sources of identiﬁcation for the model remain unchanged as well.

4.2. EFFECT OF GENERAL CONTROLS
The initial estimates of the probability a single woman is in poverty are presented
in the center columns of Tables A-6 through A-8. A quick summary of the results for the

economic and demographic controls will be followed by discussion of the EIT C’s

88

estimated impact. The effects of many factors on poverty are predictable. For example,
older women are less likely to be poor, although the effect diminishes with increased age.

Higher educational attainment lowers the probability of being in poverty. The
probability that women without a hi gh-school diploma are in poverty is consistently
estimated to be more than 20 percentage points higher than the probability for women
with a high-school diploma, other factors unchanged. All of the estimates are strongly
signiﬁcant. Education beyond high school is, as would be expected, estimated to reduce
the probability of being in poverty. Some college education reduces the probability by an
estimated 4.7 percentage points, while the completion of a bachelor’s degree or more
lowers the probability by an additional 8.1 percentage points.

To assess the impact that completing lower levels of education have on poverty
status, interaction terms were included for nonwhites with low educational attainment,
deﬁned as no more than high school, and women with children who also completed no
more than a hi gh-school education. The effects on nonwhites are discussed below. For
women with children, the estimated coefﬁcients are positive consistently and signiﬁcant.
This implies that there is a correlation between motherhood, combined with lower
educational attainment, and an increased probability of poverty. This effect is relatively
strong, controlling for other factors.

Whites have a lower probability of being poor than nonwhites. Nonwhites are
estimated to be 5.8 percentage points more likely to be in poverty than whites, controlling
for other factors. This effect increases for nonwhites who have completed relatively low
levels of education. The completion of no more than a hi gh-school education is

associated with a 3.7-percentage-point increase in the probability of poverty for nonwhite

89

women, for a cumulative effect of 9.5-percentage-point increase in the probability that a
nonwhite woman with no more than a high school education in is poverty. These
estimates are statistically signiﬁcant at traditional levels.

Single mothers, especially those with small children (under age 6), are much more
likely to be poor. The evidence here implies that the probability a single woman is in
poverty is approximately 7.1 percentage points higher on average for each additional
child she has compared to a woman without children. For women with small children,
the estimates from this sample imply that, on average, a single woman with a child under
6 years old is 14 percentage points more likely to be in poverty than a woman without
children.

Higher unemployment rates have a strong impact on the poverty rate. The
average effect of higher unemployment is estimated to be approximately 0.7, with a one
percentage point increase in the unemployment rate associated with an increase in the
probability of being below poverty of almost 0.7 percentage points. To assess whether
higher unemployment affects single mothers more directly, an interaction term between
the unemployment rate and having any children was included. The estimated coefﬁcient
for the interaction term is generally positive, indicating some correlation between higher
state unemployment and increased poverty among single mothers, but the estimates are
not statistically signiﬁcant. These estimates indicate that the overall effect of higher
unemployment is felt broadly among all single women, with some indication that the
effects might be slightly more acute for single mothers.

The estimates for the effect of higher unemployment rates on poverty are sensitive

to the speciﬁcation of the model. For example, the estimated effect increases by

90

approximately one-third (to more than 0.9) using the subsample of women with lower
educational attainment (Tables A-9 through A-l 1). It seems clear that economic
conditions have a signiﬁcant impact on poverty status among single women, both those
with and those without children.

The estimated effect of a higher real value of monthly welfare beneﬁts generally
indicates that higher monthly beneﬁts reduce poverty, although the estimates are not
signiﬁcant. The estimated coefﬁcients are small, negative, and have standard errors
much larger than the coefﬁcient estimates. This indicates a mild correlation between
higher levels of public assistance and reduced poverty.

However, when the sample is restricted to women with lower levels of education,
all of the estimated coefﬁcients are positive, although the estimates are not signiﬁcant at
traditional levels. The coefﬁcient estimates are of similar magnitude to the standard
errors, so these estimates are closer to statistical signiﬁcance. These results point to a
stronger correlation between higher beneﬁt levels and increased poverty among women
with reduced educational attainment. This evidence is consistent with the concern that
higher public-assistance beneﬁts encourage recipients to withdraw from the labor market.
As a result, there is an increase in poverty. The expansions of the EIT C were intended to

combine assistance with work, helping to move families toward self-sufﬁciency.

4.3. EFFECT OF THE EITC
Basic Model
Increases in the federal earned income credit are correlated with decreases in

poverty, as is summarized in Table 4.1. This ﬁnding is consistently replicated in most of

91

the speciﬁcations tested. This seems to conﬁrm that using the EITC to encourage labor-
force participation is an effective way to reduce poverty. This is consistent with the
general evidence discussed above and in previous literature that the EITC has contributed
to increased labor-market participation among single mothers, especially those with
lower educational achievement. The estimates imply a decrease in the probability a
single mother is in poverty of between 3.6 and 4.9 percentage points on average due to
the EIT C after the 1993 expansion, compared to the estimated effect with no EITC over
the entire sample period.

The estimated effect of the EITC is attributable only to the federal EITC in the
basic model. This is evident because the estimates for the federal EITC in panel A of
Table 4.1 are larger in absolute value than the estimated coefﬁcient for the combined rate.
Each control variable for the EITC is signiﬁcant at traditional levels. The estimated
coefﬁcients for the combined and federal subsidy rates need to be multiplied by the
existing subsidy rate to evaluate the marginal effect. For example, an increase in the
wage subsidy rate of 0.2, which is similar to the magnitude of the 1993 expansion, would,
on average, lower the probability of a single mother being in poverty by 2.4 percentage
points. Based on estimates presented in Table 4.1, the probability of poverty for a single
mother is 3.6 to 4.9 percentage points lower due to the EITC.

In the basic model, state EITCs are not associated with reduced poverty. In fact,
the estimated coefﬁcients for state credits are positive. While these estimates are not
signiﬁcant, it would seem a little strange for state EITCs to have the opposite effect on
poverty from that estimated for the federal EITC. This result is directly at odds with

those reported by Neumark and Wascher (2001).

92

As with the estimates for participation presented above, year/child interactions
were added to the basic model in order to better identify the effects of state EITCs. The
estimates, presented in panel B of Table 4.1, are qualitatively similar to the results in
panel A. The estimated coefﬁcients for the combined federal/state variable and the
separate federal variables are all negative, indicating a correlation between increases in
the EITC and reductions in poverty. However, the estimated coefﬁcients are not
signiﬁcant when the year/child interactions are included. This is again due primarily to
the collinear nature of the federal EITC variables and the year/child interactions. The
inclusion of the year/child interactions has no effect on the estimated effect of state
EITCs. The coefﬁcient estimates are slightly smaller, but they remain positive and
insigniﬁcant, indicating no correlation between state EITCs and reduced poverty.

It should be noted that both federal and state EITC payments are excluded when
determining poverty status. Including these payments in income would certainly result in
some reduction in measured poverty. To gain some insight as to the additional reductions
in poverty that would occur, if EITC beneﬁts were included in income when determining
poverty status, an estimate of the amount, if any, which each woman in the sample would
receive was calculated. This EITC amount was then added to family income. This new
measure of family income was then compared to the corresponding poverty threshold,
based on family size.

As expected, including EITC beneﬁts results in lower poverty among women with
children. When the EITC is included in income, the estimated poverty rate among

women with children is reduced by approximately 2 percentage points in 1992 and 1993,

93

below the totals reported in Table 3.2. The difference increases to 4.5 percentage points
in 1995 and 1996.

Probit estimates were then obtained using income including the EITC, relative to
the poverty threshold. As the above paragraph might suggest, the estimated coefﬁcients
for the EITC variables increase substantially (in absolute value). The estimated partial
derivative for the combined federal and state wage-subsidy rate is -0.179, compared with
0106 presented in Table 4.1. This implies that, with EITC beneﬁts included in income,
the expansion of the EITC in 1993 resulted in a decrease in poverty of approximately 1.5
percentage points over and above the estimate presented in Table 4.1.

As might be expected, the estimated effect of including EITC beneﬁts in income
is even larger when the sample is limited to women with no more than a high-school
education. The estimated coefﬁcient for the combined federal and state wage-subsidy
rate, as well as the estimated coefﬁcient for the federal wage-subsidy rate alone, both
increase in absolute value by more than 0.1, from the estimates presented in Table 4.1.
The estimates are in the range of -0.33, implying that the EITC expansion in 1993
lowered poverty among single mothers with relatively low levels of education by

approximately 6 percentage points, when EITC beneﬁts are included in income.

High School Education or Less

The estimated impact of the federal EITC on the probability of poverty increases
signiﬁcantly in absolute value when the sample is limited to women with no more than a
high-school education. The estimates presented in panel C of Table 4.1 almost double

from those in the ﬁrst panel, indicating that the EITC has a larger effect on the subgroup

94

of the population with lower educational attainment. In every case, the estimates for the
federal EITC indicate that the credit was associated with larger reductions in poverty than
the estimates obtained using the entire sample. The estimates were also signiﬁcant at
higher conﬁdence levels. This is consistent with expectations a priori, since women with
a high-school diploma or less are more likely to be in the population affected by the
credit, and thus the probability is higher that expansions in the EITC affect the poverty
status of these women.

When the basic model including the year/child interactions is estimated using the
low-education subsample, none of the EITC variables has estimated coefﬁcients that are
signiﬁcant at any traditional conﬁdence level. This is most likely due to the strong
collinearity between the year/child interaction terms and the federal EITC. However, the
coefﬁcient estimates for the year/child interactions are negative and approach statistical
signiﬁcance for years after 1988. This is the best example of the EITC’s impact being
captured by the year/child interactions. When the trend variables for children are added

later, the EITC variables regain their magnitude and statistical signiﬁcance.

New York Excluded

It is possible that state EITCs are not found to have signiﬁcant effects on poverty
due to the poor performance of the New York, similar to the effects discussed above for
labor supply. To address this possibility, the observations from New York were
excluded. Estimates were obtained using the remaining states, including either all state
EITCs or only states with refundable credits. These estimates are presented in panel B of

Table 4.1.

95

State-level credits continue to be correlated with higher levels of poverty,
controlling for other factors, even with New York excluded. The estimated coefﬁcients
become smaller when compared with the basic model, and the estimates remain
statistically insigniﬁcant. The results are qualitatively the same, regardless of whether

the state EITC variable includes all state EITCs, or just refundable EITCs.

Welfare Reform

The next speciﬁcation controls for the effect of welfare reform. The welfare
system in existence prior to the enactment of the PRWORA created incentives that did
not foster independence. The accelerating reform trend, reﬂected in the fact that the
majority of states acquired waivers, was evidence of growing dissatisfaction with the
existing system.

The emphasis that many states placed on welfare reform has the potential to bias
the estimates of the impact of EITCs. If the inﬂuence of welfare reform is omitted, the
EITC may capture some of that inﬂuence, resulting in biased estimates. In order to
control for this inﬂuence, the qualitative measure of the intensity of welfare reform
constructed by Ellwood was used.

The estimated coefﬁcients are presented in panel F of Table 4.1. From the ﬁrst
panel of Table 4.1, the estimated coefﬁcients for the combined federal and state subsidy
rate was —0.106. When the welfare reform variable is added, the estimated coefﬁcients
for the combined EITC rate remains essentially unchanged. Similar results are obtained
using the separate federal EITC. All the estimates for the federal EITC remain

statistically signiﬁcant. Rather than reducing the estimated effect of the EITC materially,

96

the reduction in poverty attributable to the EITC remains stable. The state EITC estimate

remains positive and insigniﬁcant.

Trends for Women with One Child or Two or More Children

Given that the inclusion of year/child interactions resulted in insigniﬁcant
estimates of the EITC’s effect on poverty, and that the year/child interactions are
collinear with the federal EITC controls, an alternative method of controlling for
additional inﬂuences on the poverty status of single mothers over time other than the
EITC was desirable. To accomplish this, trends were added to the basic model for
women with one child and women with two or more children. These trend variables will
capture the inﬂuence of factors that are not previously included in the model that affect
one of the subsets of single mothers.

The estimates including the trend variables, presented in panel G of Table 4.1, are
qualitatively unchanged from the basic model. All of the estimates in panel G are slightly
larger in absolute value than the estimates in panel A, and the combined subsidy rate and
the federal EITC rates are estimated to be strongly signiﬁcant. The consistency of these
estimates across speciﬁcations provide strong evidence of the efﬁcacy of the federal
EITC as a poverty reduction tool. The estimates for state EITCs remain positive and

insigniﬁcant.

Logged E1 T C Effects

In Chapter 3, evidence was presented to support the view that the EIT C affects

labor supply with a lag. Speciﬁcally, the estimates indicate that the federal EITC may

97

signiﬁcantly increase participation with a two-year lag. Similar estimates were obtained
for the impact of the EITC on poverty. These results are presented in Table 4.2.

As a whole, the estimated effects of federal and state EITCs on reducing poverty
consistently increase over time. The largest reductions occur after two years, the same
length of time indicated by the estimates on participation. While most of the estimates
are not statistically signiﬁcant, all of the estimated coefﬁcients on the lagged federal
EITC variables are negative. When the state EITC variable is limited to refundable
credits, the estimated coefﬁcient on the two-year lag is negative and signiﬁcant at the ten
percent level. This may indicate that state EITCs do help reduce poverty, but that the

impact takes two years.

4.4. COMPARISION WITH MINIMUM WAGES AND WELFARE WAIVERS

The EITC is just one of the policies used to encourage single mothers to work and
then support them in the labor market. As discussed above, welfare policy may be
altered in ways other than waivers of existing federal regulations. In addition, minimum
wages are often cited as an appropriate policy to help low-income families.

In the results obtained here, minimum wages are not associated with reduced
poverty. The estimated coefﬁcients, summarized in Table 3.7, are generally positive,
implying a correlation between higher minimum wages and increased poverty. None of
the estimated coefﬁcients are signiﬁcant, with test statistics well below 1. The estimates
obtained using the low-education subsample become larger in magnitude and are closer

to statistical signiﬁcance. The estimates for this subsample indicate that among the

98

women most likely to be near poverty, higher minimum wages are at least weakly
correlated with higher poverty.

These results pointing to the ineffectiveness of the minimum wage are consistent
with two ﬁndings in the literature. Burkhauser et al. (1996) found that a large percentage
of the workers receiving the minimum wage are not poor, leading to a “ﬁrzzy”
relationship between wage rates and economic well-being. Second, Neumark and
Wascher (1998) found that higher minimum wages result in counter-balancing ﬂows into
and out of poverty. Increases in the legal minimum wage allow families with workers at
the new minimum to escape poverty, but some families formerly above poverty fall back
below the poverty level. This follows from the standard model of the minimum wage.
Those workers employed at the new minimum may receive an increase in pay, but those
workers who are either laid off or involuntarily underemployed see their incomes fall. In
light of the evidence presented here that higher minimum wages are not correlated with
reduced poverty, and the strong evidence that the expansions of the EITC are correlated
with reduced poverty among single mothers, it would appear that the EIT C was a more
effective anti-poverty tool during the 19905.

The implementation of a welfare waiver is associated with reductions in poverty.
As a reminder, waivers were controlled for using a variable equal to the fraction of each
year a state had a statewide waiver in place, interacted with an indicator variable for
children. So the waiver variable would be equal to one if a single woman with at least
one child lived in a state that had a waiver in place for the entire year.

The estimated effect of welfare waivers on poverty is negative, signifying a

correlation with reduced poverty. As shown in Table 3.7, the estimated coefﬁcients

99

range from -0.014 to -0.020, with the largest estimates (in absolute value) obtained when
the sample is restricted to women with no more than a high school education and when
controlling for the aggressiveness of welfare reform. The estimates are generally
signiﬁcant. This would seem to imply that welfare waivers were somewhat effective in
reducing poverty, with the average effect around 1.5 percentage points. This result is
consistent with the results of Schoeni and Blank (2000), who also found that waivers
reduced poverty signiﬁcantly. However, the model used by Schoeni and Blank excludes
controls for the expansions of the EITC, and the results here indicate the EITC had a
signiﬁcant inﬂuence both on participation and reduced poverty. Excluding the EITC
would bias the estimated impact of waivers upward. The 1993 expansion of the EITC
occurred during the acceleration of states receiving waivers. Thus, omitting the
signiﬁcant effect of the EITC would bias the estimate.

To examine the extent to which excluding the EITC variables do bias the
estimated effect of welfare waivers, regressions were estimated without any EITC
controls. The results, which appear in Table 3.8, indicate that having a waiver in place
for an entire year would reduce poverty by 3.4 percentage points, more than double the
estimate of 1.6 percentage points from panel A of Table 3.7. This would seem to indicate
that omitting the EITC results in a signiﬁcant bias in the estimated effectiveness of
welfare waivers.

Compared with the expansions of the EITC, welfare waivers are correlated with
somewhat smaller changes in the probability that single mothers would be in poverty.
The estimated effect of the EITC on poverty is also understated in these estimates

because the dollar value of the EITC, which has increased for all eligible recipients over

100

time, has been excluded from income. When EITC beneﬁts are included, the total impact
of the EITC on reducing poverty among single mothers is signiﬁcantly larger. This
provides additional evidence that the EITC has been more effective at reducing poverty

than either increases in the minimum wage or the enactment of welfare waivers.

4.5. ESTIMATES FOR 150 PERCENT OF POVERTY

The framework used thus far was once again employed to estimate the probability
a woman would be near poverty, deﬁned as at or below 150 percent of the poverty level.
The results for near-poverty status are qualitatively similar to the results for poverty
discussed above. As such, the discussion here will focus on the EITC, both federal and
state credits, along with the two policy alternatives, the minimum wage, and welfare

waivers.

Effects of E] T C on near-poverty status

The estimates for near-poverty status are summarized in Table 4.3, with the full
results presented in the far-right columns of the appendix tables. In general, the estimates
indicate that the federal EITC is correlated with a decreased probability that single
mothers are below 150 percent of the poverty level. Most of the estimated coefﬁcients
are smaller in absolute value when the dependent variable is 150 percent of the poverty
level than for those obtained in the poverty estimates discussed above. The estimated
coefﬁcients for the federal EITC are negative and signiﬁcant, indicating the federal EITC

is effective at helping single mothers move above near-poverty status.

101

For the basic model, presented in the ﬁrst panel of Table 4.3, the estimate for the
separate federal EITC is -0.101. This implies approximately a 2.1-percentage-point
reduction in the probability a single mother is near poverty following the EITC expansion
in 1993, and approximately a 4.0 percentage point reduction in the probability overall due
to the EITC. The estimated reduction due to the 1993 expansion is —3.2 percentage points
using the subsample of women with low educational attainment, and —2.3 percentage
points if the New York observations are excluded.

None of the estimates for state EITCs are statistically signiﬁcant. The
coefﬁcients are generally positive, consistent with a correlation between states with
EITCs and an increased probability that single mothers in those states are below 150
percent of poverty. The estimated coefﬁcients are insigniﬁcant at all standard conﬁdence
levels. This could represent women who are in the labor force earning enough to rise
above the poverty level, but who remain close to poverty nonetheless.

Overall, the federal EITC is found to have a signiﬁcant effect on moving single
mothers above near-poverty status, deﬁned here as below 150 percent of poverty. The
estimated effect on near-poverty status is slightly smaller than the estimated effect on
poverty status. Given the signiﬁcant effects the EITC has been estimated to have on
participation and poverty, it would seem reasonable that a portion of the population has
moved into the labor force and risen above poverty. The estimates presented in this
section seem to indicate that some of these single mothers have been able to rise even
further away from poverty. The exclusion of EITC beneﬁts, both federal and state, from

the deﬁnition of income has likely reduced the measured reduction in near-poverty status.

102

If the beneﬁts of the EITC were included in income, it is likely that an additional

reduction in near-poverty status would be observed in these data.

Comparison with Minimum Wage and Welfare Waivers

The estimated impact of minimum wages is larger and more signiﬁcant when
using the higher income threshold. These results are presented in Table 3.7. The
coefﬁcient estimates are positive and signiﬁcant, with the value of the estimates around
0.22 consistently. This would imply that a 10-percentage point increase in the relative
value of the minimum wage would increase the probability a single woman is below 150
percent of poverty by 2.2 percentage points.

The estimated impact of the minimum wage increases again when the sample is
reduced to only women with a high school diploma or less. The coefﬁcient estimate of
approximately 0.4 implies a larger reduction in income for women with lower levels of
schooling when the minimum wage is increased. This ﬁnding is consistent with a shift
away from low-skilled workers, either to increased technology investment or to hi gher-
skill labor, when minimum wages are increased.

When combined with the estimates obtained for 100 percent of poverty, which
were positive and marginally signiﬁcant, there is evidence that higher minimum wages
may restrict the mobility of low-income families out of poverty or near-poverty status.
This should be compared with the estimates for the EITC, which imply a signiﬁcant
decrease in the probability that single mothers are in or near poverty. This would indicate
that, at least for women with children, the EITC is a more effective tool to reduce poverty

than the minimum wage.

103

The estimated effect of welfare waivers on near-poverty status is insigniﬁcant in
all speciﬁcations. The estimated coefﬁcients range from -0.003 to -0.009, with test
statistics well below 1.0. The evidence suggests that welfare waivers are more effective
at moving single mothers above the lower poverty threshold.

When compared with increases in the EITC, it appears that welfare waivers are
less effective in moving single mothers away from poverty. Since the broad expansion in
the use of waivers occurred later in the period covered by this study than did the ﬁrst
EITC expansions, it may take time for the effect of waivers to be as visible. Since the
implementation of the PRWORA allowed all states to reform the existing welfare system
fruther, modiﬁcations to the old welfare system may have more signiﬁcant effects after

1996.

104

Table 4.1
Estimated Effects of the EITC on Poverty

Combined
A. Basic Model
Combined Subsidy Rate -0.1057 *"‘
(0.0295)
Separate Federal/State Rates
Refundable State EITC Rate
B. High School or Less
Combined Subsidy Rate -0.2086 **
(0.0553)

Separate Federal/State Rates

Refundable State EITC Rate

C. Including Year/Child Interactions

Combined Subsidy Rate -0.0086

(0.1371)

Separate Federal/State Rates

Refundable State EITC Rate

Federal

0.1195 **
(0.0318)

-0.1219 **
(0.0316)

-0.2227 **
(0.0595)

-0.2289 **
(0.0591)

-0.0899
(0.1773)

-0.0905
(0.1775)

D. Including Year/Child Interactions - Low Education Sample

Combined Subsidy Rate 0.0742
(0.2581)
Separate F ederal/State Rates
Refundable State EITC Rate
E. Excluding New York Observations
Combined Subsidy Rate -0.1 162 *"'
(0.0317)

Separate Federal/State Rates

Refundable State EITC Rate

105

0.2589
(0.3240)

0.2549
(0.3245)

-0.1268 "-
(0.0328)

-0.1270
(0.0326)

All State

EITCs

0.0664
(0.1972)

-0.0403
(0.3673)

0.0529
(0.1980)

-0.0583
(0.3706)

0.0481
(0.2126)

Refundable
EITCs

0.1770
(0.2444)

0.1761
(0.4373)

0.1531
(0.2458)

0.1631
(0.4394)

0.1363
(0.3007)

Table 4.1 (cont'd).

F. Includes Welfare Reform

Combined Subsidy Rate -0. 1037 **
(0.0295)
Separate Federal/State Rates -0.117 **
(0.0319)
Refundable State EITC Rate ~0.l 194 **
(0.0316)

G. Basic Model using Trends for l and 2 or More Children

Combined Subsidy Rate -0. 1095 '“'
(0.0362)
Separate Federal/State Rates -0. 1288 *"'
(0.0390)
Refundable State EITC Rate -0.l321 “"
(0.0390)

** Indicates signiﬁcant at 5% conﬁdence level.
"' Indicates signiﬁcant at 10% conﬁdence level.

106

0.0614
(0.1971)

0.0691
(0.1971)

0.1685
(0.2441)

0.1845
(0.2450)

Table 4.2
Estimated Effects of the EITC on Poverty
Using Lagged EITC Controls

All State Refundable
Combined Federal EITCs EITCs
A. Including Concurrent and Two Lagged Rates
1. Concurrent 0.0085
(0.1920)
Lagged One Year -0.0243
(0.3005)
Lagged Two Years -0.1575
(0.1924)
2. Concurrent -0.0407 0.5188
(0.2068) (0.7698)
Lagged One Year -0.0198 0.1294
(0.3260) (1.1139)
Lagged Two Years -0.1128 -0.8454
(0.2056) (0.7342)
3. Concurrent —0.0383 0.4119
(0.2068) (0.9290)
Lagged One Year -0.0314 1.0184
(0.3257) (1.5422)
Lagged Two Years -0.1065 -l.7408 *
(0.2058) (1.0805)
B. Including Two Legged Rates
1. Lagged One Year -0.0122
(0.1304)
Lagged Two Years -0. 1607
(0.1780)
2. Lagged One Year -0.0750 0.7971
(0.1399) (0.5494)
Lagged Two Years -0.0963 -0.9800
(0.1882) (0.7087)
3. Lagged One Year -0.0860 1.6158 "”"
(0.1399) (0.7886)
Lagged Two Years -0.0885 -1.9306 **
(0.1884) (1.0066)

** Indicates signiﬁcant at 5% conﬁdence level.
* Indicates signiﬁcant at 10% conﬁdence level.

107

Table 4.3
Estimated Effects of the EITC on 150 Percent of Poverty

All State Refundable
Combined Federal EITCs EITCs
A. Basic Model
Combined Subsidy Rate -0. 1012 **
(0.0374)
Separate Federal/State Rates -0.1133 ** 0.0521
(0.0403) (0.2690)
Refundable State EITC Rate -0.1149 ** -0.1371
(0.0399) (0.3318)
B. High School or Less
Combined Subsidy Rate -0. 1561 **
(0.0062)
Separate Federal/State Rates -0.1975 ** 0.3584
(0.0668) (0.4673)
Refundable State EITC Rate -0.1884 ** 0.2012
(0.0658) (0.5006)
C. Including Year/Child Interactions
Combined Subsidy Rate -0.0685
(0.1838)
Separate Federal/State Rates -0.2247 0.0503
(0.2255) (0.2696)
Refundable State EITC Rate -0.2252 0.1411
(0.2258) (0.3344)
D. Including Year/Child Interactions - Low Education Sample
Combined Subsidy Rate 0.2512
(0.3171)
Separate Federal/State Rates 0.1265 0.3444
(0.3590) (0.4718)
Refundable State EITC Rate 0.1353 0.1909
(0.3600) (0.5059)
E. Excluding New York Observations
Combined Subsidy Rate -0.1133 "
(0.0402)
Separate Federal/State Rates -0. 1223 ** 0.0280
(0.0416) (0.2896)
Refundable State EITC Rate -0. 1221 0.0468
(0.0412) (0.4075)

108

Table 4.3 (cont'd).

F. Includes Welfare Reform

Combined Subsidy Rate -0.0975 **
(0.0375)
Separate Federal/State Rates -0.1086 **
(0.0404)
Refundable State EITC Rate -0.1101 **
(0.0400)
G. Basic Model using Trends for l and 2 or More Children
Combined Subsidy Rate -0.1376 "
(0.0462)
Separate Federal/State Rates -0.1592 **
(0.0497)
Refundable State EITC Rate -0.1623 **
(0.0495)

" Indicates signiﬁcant at 5% conﬁdence level.
"' Indicates signiﬁcant at 10% conﬁdence level.

109

0.0426
(0.2687)

0.0651
(0.2688)

0.1200
(0.3310)

0.1774
(0.3331)

CHAPTER 5

COMPARISON TO PREVIOUS RESEARCH AND CONCLUSIONS

5.1. COMPARISON TO PREVIOUS RESEARCH

The results are consistent with previous research for the EITC’s effect on labor-
market participation. There are three thorough econometric studies of the effects on the
labor-market participation of single mothers. These are Eissa and Liebman (1996),
Meyer and Rosenbaum (2001), and Ellwood (2000). All of these found the EITC has a
positive effect on labor-market participation. Neumark and Wascher (2001) focus more
on poverty status, although they also report some evidence that the federal EITC has a
signiﬁcant effect on labor-force participation among CPS families. This study
corroborates those ﬁndings, while expanding the set of control variables to include
interstate variation in minimum wages, state-level EITCs, and welfare policy. While the
EITC is correlated with increases in labor-market participation, minimum wages and
welfare waivers do not consistently produce increases in participation.

Eissa and Liebman focused on the period prior to OBRA 1990, so they only
examined the effect of the expansion due to the Tax Reform Act of 1986. The impact on
participation measured here is consistent with the lower end of their results. In the Eissa
and Liebman study, the effect on participation for women with less than a high school
education is 6.1 percentage points. The largest estimates presented here imply a marginal
response of 4.8 percentage points to the 1986 expansion. The lower estimate is due to the
older sample used for this study. Eissa and Liebman use a sample of single women ages

16 to 44. The sample here includes single women ages 25 to 50. Older mothers appear

110

to have increased labor-force participation at a lower, but still signiﬁcant, rate. The
results here build on the Eissa and Liebman study by extending the period of time to
encompass the 1990 and 1993 expansions of the EITC as well. The results indicate that
the effect on participation is larger as the EITC has increased, with the 1993 expansion
resulting in an increase in participation of nearly 5 percentage points among women with
a high school education or less.

The results are broadly consistent with the other studies as well. The addition of
explanatory variables controlling for the real minimum wage does little to change the
qualitative results found by Meyer and Rosenbaum (2000 and 2001) or Ellwood (2000).

In general, the estimated effect of the EITC is smaller, as might be expected in a
sample of older women. In the previous studies by Meyer and Rosenbaum, larger
participation effects were found than those reported here. Meyer and Rosenbaum include
younger women than are included here. If the broader sample of single mothers exhibit
larger participation effects than those for the older sample used here, it is reasonable to
suppose that the poverty-reducing effects of the EITC reported here may be biased
downward.

To test the hypothesis that a younger sample would produce larger estimated
effects on participation, a sample of single women ages 18 to 44 was obtained from the
March CPS for the years 1983 to 1996. The estimated effects of the EITC on labor-
market participation among this younger sample are presented in Table 5.1. When
compared with the results presented in Table 3.3, the estimated coefﬁcients for the
federal EITC are slightly larger for the basic model including all single women. The

estimated coefﬁcient on the combined federal/state EITC for the younger sample is

111

0.118, compared to 0.102 for the sample used for most of this study. Similar differences
are found for the separate federal EITC. Larger differences occur when the sample is
limited to women with no more than a high-school education. These provide evidence
that at least some of the difference between the participation effect measured by Meyer
and Rosenbaum and the estimate presented here is due to the older sample used here.

Ellwood’s study also attempts to control for the interstate variation in the relative
intensity of moving residents off of public assistance. The addition of a measure of
welfare-reform aggressiveness to this study results in relatively small decreases in the
measured effectiveness of the EITC, both for increasing participation and reducing
poverty. Welfare reform itself is estimated to have small but statistically signiﬁcant
effects on participation and poverty.

This study is the second to attempt to measure the effects of the EITC on the
poverty status of single mothers, using a different methodology than Neumark and
Wascher (2001). Each expansion of the federal EITC is associated with decreases in
poverty. The key difference between the ﬁnding here and those of Neumark and
Wascher is the effect of state EITCs.

Neumark and Wascher ﬁnd that the enactment of state EITCs signiﬁcantly
increases the transition of families out of poverty. The impact of the federal EITC is
insigniﬁcant and generally estimated to be negative, indicating a correlation between the
expansion of the federal EITC and increased movement into poverty. Neumark and
Wascher attribute the seeming inconsistency of the results for the federal EITC to poor

identiﬁcation. In their speciﬁcation, state EITCs are better identiﬁed than the federal

112

EITC. The signiﬁcant results for state EITCs are seen to provide general evidence of the
efﬁcacy of EITCs at reducing poverty.

Neumark and Wascher use year/child interactions in their estimation procedures
to control for factors affecting families with children that are not otherwise explicitly
contained in their model. When these year/child interactions are used here, the effect of
the EIT C on poverty (but not participation) becomes insigniﬁcant. Since that is unique to
year/child interactions among the various speciﬁcations used here, it is likely that the
estimates using year/child interactions are adversely affected by the collinearity between
the interaction terms and the federal EITC controls. In the remaining speciﬁcations,
including estimates obtained using trend variables to capture inﬂuences affecting single
mothers over time, the federal EITC is found to signiﬁcantly reduce poverty among
single mothers. State EITCs are not found to signiﬁcantly reduce poverty in any of the
estimates presented here.

The signiﬁcant poverty reductions reported here are contrary to Browning’s
(1995) predicted effects of the OBRA-1993 expansion. Browning found that the EITC
was likely to reduce the disposable income of many recipients. The driving force for this
decline was the steep increase in the phase-out rate due to the 1993 expansion of the
EITC. The continued decline in poverty after 1993 would seem to contradict this
prediction.

However, Browning’s estimated overall effects are not refuted by this study. The
predicted effects in Browning’s paper are driven by labor supply reductions among those
already working. The segment of the population most likely to experience these

reductions is married couples. In fact, Eissa and Hoynes found that married women did

113

reduce their labor supply in response to the 1993 EITC expansion. This study covers
single mothers. While single mothers are estimated to increase their participation,
married families in the phase-out region may reduce their collective labor supply, either
through reductions in hours or withdrawal from the labor market altogether. Reductions
in labor supply by families in the phase-out range would mitigate the EITC’s overall
effectiveness.

The evidence presented here on the EITC’s effect on poverty can be contrasted
with the lack of evidence supporting similar effects from increasing the minimum wage.
The evidence on minimum wages presented here indicates that the minimum wage does
not have signiﬁcant effects on either labor-force participation or poverty. These ﬁndings
are consistent with Neumark and Wascher (1998). Their results indicate that minimum-
wage increases are associated with movement both into and out of poverty, resulting in
an insigniﬁcant overall effect. This is consistent with some reductions in hours worked
among low-wage workers. Because the EITC does not affect labor demand negatively,
there is no offsetting disemployment effect.

Schoeni and Blank (2000) report signiﬁcant increases in participation and
reductions in poverty due to welfare waivers in advance of the PRWORA. Those
ﬁndings are only supported here for poverty when controlling for the inﬂuence of the
EITC. When the EITC variables are excluded item the estimation model used here, as
was done in the Schoeni and Blank study, there is evidence that excluding the EITC
introduces a signiﬁcant bias into the estimated effectiveness of waivers. The signiﬁcant
effects of the EITC on poverty and participation remain robust, regardless of whether

changes in welfare policy are modeled using waivers or the qualitative measure of reform

114

intensity created by Ellwood. All of the estimates reveal a strong effect for the EITC,

moving single mothers into employment and out of poverty.

5.2. CONCLUSION

In general, the results indicate that the federal EITC has led to both increased
labor-market participation and reduced poverty among single adult mothers. After the
1993 expansion of the EITC, labor-market participation has increased by between 3.5 and
4.9 percentage points due to the EITC. For women with a high school diploma or less
education, the effect is approximately twice as large. Women with lower levels of
education also increased their participation by signiﬁcant amounts following the earlier
EITC expansion in 1986.

The EITC is also found to have contributed to a signiﬁcant decline in poverty
among single women with children. The estimated impact of the EITC on the poverty
status of women with no more than a high school education is even larger than the
estimates for the entire population. The estimated reduction in poverty among single
mothers due to the 1993 expansion of the EITC is approximately 1.9 percentage points.

With increased participation and reduced poverty rates due to the expansions of
the EITC, it is no surprise that the EITC is also estimated to have reduced the likelihood a
single woman with children is below 150 percent of the poverty level. In general, the
estimated effects indicate that the EITC has a smaller impact in moving women away
from near-poverty status than away from poverty, although the difference is small.

The estimated impacts of the EITC on participation and poverty are relatively

consistent across multiple speciﬁcations. The inclusion of minimum-wage controls,

115

welfare-beneﬁt rates, welfare-reform controls such as waivers, and state-level EITCs
leave the basic conclusions unchanged. The older sample used for this study displays a
reduced responsiveness to the expansions of the EITC, relative to the estimates obtained
using samples that included younger women. Given the larger estimates of increased
participation reported using the younger samples, the results presented here may
understate the EITC’s overall impact on poverty.

Some evidence is presented to indicate that refundable state-level EITCs are
marginally effective at increasing labor-force participation. There is also evidence that
the full impact of expanding the EITC is not felt for at least two years. The lagged effect
was found for both labor-force participation and poverty. There is also some evidence
that the overall estimates for state credits are obscured by the New York EITC, enacted in
1994.

Overall, expansions of EITC have increased labor-force participation and reduced
poverty among single mothers. Based on the results obtained here, neither the minimum
wage nor changes in welfare policy produce similar results. As a result, the EITC should
be considered a useful tool in providing assistance to low-income families, encouraging

these families to enter the labor market, and then assisting them to rise out of poverty.

116

Table 5.1

Estimated Effects of the EITC on Labor Supply

Using Sample of Single Women Ages 18 - 44

Combined
A. Basic Model
Combined Subsidy Rate 0.1182 **
(0.0178)
Separate Federal/State Rates
Refundable State Rate
B. High School or Less
Combined Subsidy Rate 0.2836 *"'
(0.0378)

Separate Federal/State Rates

Refundable State Rate

C. Including Year/Child Interactions

Combined Subsidy Rate -0.0289

(0.0641)

Separate Federal/State Rates

Refundable State Rate

"“" Indicates signiﬁcant at 5% conﬁdence level.
* Indicates signiﬁcant at 10% conﬁdence level.

117

Federal

0.1377 **
(0.0193)

0.1294 "
(0.0193)

0.3261 **
(0.0408)

0.3136 "
(0.0409)

0.0768
(0.0986)

0.0719
(0.0985)

All State

EITCs

-00973
(0.0866)

-0.1896
(0.1804)

-0. 1071
(0.0865)

Refundable
EITCs

0.1371
(0.1345)

0.1751
(0.2729)

0.1120
(0.1343)

CHAPTER 6

EFFECT OF THE EITC ON HUMAN-CAPITAL ACCUMULATION

6.1. INTRODUCTION

The earned income tax credit (EITC) is designed to encourage parents to increase
their labor-market participation by providing a wage subsidy of up to 40 percent. The
EITC is eventually phased out, resulting in relatively high marginal tax rates that may
discourage additional labor earnings. After undergoing three expansions between 1985
and 1995, the EITC has become a key component of policy toward the poor and near-
poor populations in the United States.

The wage subsidy portion of the EITC is designed to encourage parents who are
either outside the labor force, or marginal participants, to increase their labor supply.
While the credit is available to all parents who meet the eligibility criteria, the EITC is
often thought of as a program for single mothers. Indeed, most of the research on the
EITC has focused on the program’s effects on single women with children.

Increased labor supply among single mothers may be seen as a measure of the
ElTC’s effectiveness. However, it is possible that, especially for a young parent,
additional labor supply will reduce the time available for education. This is not
necessarily bad, but it does represent a choice of a particular type of human capital
development and accumulation (work experience as opposed to education) that may have
long-term ramiﬁcations on the woman’s income, poverty status, and economic well—
being. It is ambiguous, a priori, whether additional substitution of work experience for

schooling makes a young, single parent better or worse off, but the EITC clearly

118

subsidizes labor-market participation by single parents. This is especially true for young
women, because the probability of being enrolled in school declines with age.

In addition to the time constraint, the high marginal tax rates associated with the
phase out of the EITC may reduce the incentive that an EITC recipient has to participate
in additional training. The marginal tax rate for an EITC recipient may reach more than
45 percent, and may last over a range of earnings exceeding $15,000. Such high tax rates
may discourage additional education for EITC recipients at the margin. The overall
effect should depend on the expected wage after the additional education is completed.
High tax rates may discourage training that results in an expected future wage, and
resulting total labor income, within the EITC phase-out range. However, if the expected
net wage after completing additional training is high enough to allow the worker to move
beyond the phase-out range, the EITC’s impact on the net wage should not affect training
decisions.

The EITC may provide the additional income necessary for single parents to pay
for additional training and schooling. Since most EITC recipients receive the credit as a
tax refund, the EITC may aid some recipients in overcoming possible liquidity
constraints in the market for training. The llunp-sum payment that recipients receive may
provide the necessary funds to pay tuition bills.

Taken together, the impact of the EITC on additional schooling and training is
ambiguous. The wage subsidy and high marginal tax rates may discourage additional
schooling. However, the additional income may allow a recipient to invest in additional

training.

119

Previous analyses of the earned income tax credit have focused on the short-term
impact, i.e., the impact on labor supply. Does the credit encourage labor-market
participation among single mothers? Does the credit reduce hours worked by those
already participating in the labor market? This paper begins to look at one of the
potential longer-term impacts. Does the credit affect the accumulation of human capital?
Speciﬁcally, is there evidence that the earned income tax credit shifts women with
children away from schooling and toward the labor market, resulting in the accumulation

of work experience and a reduction in schooling?

6.2. DESCRIPTION OF STUDY

The EITC provides beneﬁts predominantly to families with qualifying children
who earn relatively low wages. Beneﬁts from the program have increased signiﬁcantly
over time, taking the program from a small tax credit with relatively few recipients to a
major expenditure program that exceeds all direct expenditure programs targeted at the
working poor.

This paper will use data from the March Current Population Surveys to examine
whether the expansions in the EITC have resulted in reduced enrolhnents and increased
labor supply for young, single mothers, relative to their contemporaries without children.
The study will examine the effects of the EITC by controlling for other factors that might
affect labor-market participation by young adults, such as increases in the minimum wage
and economic growth on the state level.

Evidence has accumulated that the EITC has increased the probability that young

mothers work, relative to their childless peers. A number of studies have found that the

120

expansions of the EITC since 1986 have resulted in increased labor-market participation
among single mothers. These studies were discussed in Chapter 2. It is likely that some
of this increase has taken place among women who were still in the prime school years,
say, younger than 25.

Because of the demands of parenting, it is possible that increased labor-market
activity would reduce the amount of time available for schooling. Both work experience
and schooling increase the lifetime earnings proﬁle relative to not working or going to
school. However, the EITC changes the current set of relative prices, and would change
the lifetime discounted present value of the choice of work or school. Given this shift in
relative prices, it might be expected that the probability that young mothers remain in
school would decline, while the probability that young mothers work would increase.

Human-capital theory describes two primary means of accumulating human
capital: schooling and on-the-job training. These have been discussed thoroughly by
Becker (1993), Mincer (1974), and Rosen (1977). Wages are generally viewed in this
literature as a return to accumulated human capital, reﬂecting the higher marginal product
of workers with larger stocks of human capital. Human capital can be accumulated by
investments in schooling (used here to refer to formal primary, secondary, post—
secondary, or vocational education), or through investments in on-the—job training. While
both types of training result in the accumulation of additional human capital, they do not
represent identical investments. Schooling often represents general human-capital
investments that are valuable at a large number of ﬁrms, often in a range of speciﬁc jobs.
On—the-job training may tend to be less general, with the majority of its beneﬁts accruing

at the current job, although this is not required.

121

General training applies broadly across many ﬁrms and jobs. Because of the
possibility for employees to change jobs after receiving general training, ﬁrms require
employees to pay the cost of general training. Employees can pay the costs explicitly
out-of-pocket, or implicitly through lower wages, if their employer provides the general
training. Speciﬁc training is deﬁned to be ﬁrm speciﬁc, and provides increased
productivity at the current job, but is of little or no use to competing employers.
Employers will be willing to share the cost of this training with their employees. In
exchange for sharing the cost of training, part of the increased productivity is also shared.
Sharing the costs and beneﬁts protects the ﬁrm from the worker resigning immediately
after the completion of training, by giving the worker an incentive to stay, a higher wage.

Barron, Berger, and Black (1999a, 1999b) present evidence indicating that most
training is not general, but rather speciﬁc in nature. Their evidence points to small
variations in wages due to training. Workers requiring less training are generally paid a
higher starting wage, while starting wages do not fall for workers requiring more training.
Rather, ﬁrms pay for training and recover the cost through increased productivity. This
implies that most training is speciﬁc in nature, preventing workers from capturing a
higher return (wage) from a competing ﬁrm.

While most training may be speciﬁc, as discussed above, employers would
require a worker to pay for any general training the worker receives because such
training, once received, allows the worker to seek a higher wage from competing
employers. The ﬁrm will thus require the worker to pay for this training, either explicitly
out-of-pocket, or through lower wages. Government policies, such as minimum wages,

may hinder this type of arrangement for training. Higher minimum wages prevent lower-

122

skilled workers from paying these costs implicitly, resulting in less training being
provided by employers.

By contrast, the EITC initially provides a wage subsidy that does not affect the
labor costs of employers. If the EIT C increases the supply of low-wage workers enough,
the market wage may actually fall. This would actually lower the cost of new employees,
providing some incentive for employers to provide additional training. With a lower
market wage, employers can provide general training and effectively recoup the cost
through the lower wage, as long as the “training” wage is above the legal minimum.

There is another possible mechanism through which the EITC could increase
training. The EITC is administered ahnost exclusively between the worker and the
government through tax returns. So the EITC should not reduce offers of on-the-job
training from employers, because the cost of hiring and employing a new worker is not
increased by the EITC. For workers, the EITC effectively raises the net wage, offsetting
some or all of the cost of training the employee is paying through a lower wage. To the
extent the EITC raises the net wage above the reservation wage of some workers,
employment and training both increase. Therefore, through the EITC, it is possible that
the government can subsidize general on-the-job training, by “reimbursing” employees
through the credit.

So while an employee can acquire some general training at work, and the EITC
can subsidize some of this training for low-income workers, the employee will need to
arrange for general training that is applicable to a broad range of employers. This is
where the EITC may be an impediment to the accumulation of general training. The

phase-in range of the EITC encourages work. However, time constraints may prevent a

123

parent from both working and attending school. Increased labor-market participation due
to an expansion of the EITC would potentially imply reduced school attendance, and thus
a reduced opportunity set for general training.

As discussed in the introduction, the overall effect of the expansion of the EITC
on school enrollment is ambiguous theoretically. Any potential decline in school
attendance due to the higher wage subsidy may be offset by the increased disposable
income available to credit recipients. A goal of this study is to see if there is clear
evidence as to whether the expansion of the EITC has led to higher or lower school
enrollments among single mothers.

Preliminary evidence is presented in Tables 6.1 and 6.2. These tables compare
the combined school and employment status of single women with and without children,
for two age groups over a 10-year period. The ﬁrst group is ages 17 through 21, while
the second is ages 17 through 24.

Single women with children have steep declines in the non-enrolled/non-
employed category, while the gains are distributed mainly into employment, with single
mothers also showing gains in employed and enrolled status. The strong gains in
employment for single mothers are consistent with the EITC increasing labor-market
participation through increases in the wage-subsidy rate available in the phase-in range.
The gains in school enrollment among single mothers provide some initial evidence that
the gain in employment did not occur at the expense of additional schooling, which
increased by approximately 10 percentage points for both age groupings. For women
without children, the clearest movement over this period was from employment with no

schooling, to school with no employment.

124

6.3. RELATED LITERATURE

While the potential effects of the EITC on school enrollment have not been
examined previously, there are two areas of research related to this paper. The ﬁrst is a
rather wide body of research on the effects of minimum wages on school enrolhnent and
human-capital accumulation. In general, it has been reported that minimum wages
reduce school enrollment by teenagers (Ehrenberg and Marcus (1982) and Neumark and
Wascher (1996)). While there has been some conﬂicting evidence, the weight of the
evidence points to higher minimum wages resulting in reduced enrollment of teenagers.
There are two reasons students leave school. Higher-skill students may leave school after
minimum—wage increases to pursue work at the higher minimum wage. In addition, the
higher minimum induces some students to leave school and queue for full-time work at
the higher minimum wage.

A smaller body of literature has examined the effect of minimum wages on
training. In general, this literature has reported evidence of reduced training being
correlated with higher minimum wages (for example, see Neumark and Wascher (2000),

Hashimoto (1982), and Leighton and Mincer ( 1981)).

Minimum Wages and Enrollment

Mattila (1979) was the ﬁrst study to investigate seriously the effect of minimum
wages on youth enrollment. Using data from the October CPS, Mattila found that
displaced teenage workers do return to school. This study ﬁnds that minimum wages
increase enrollments across males and females in age groups of 16-17, 18-19, and 20-24.

Three of the six estimates were signiﬁcant, while another (females aged 20-24) was

125

NI‘L

nearly signiﬁcant. These estimates imply that young workers return to school after the
minimum wage increases, in order to increase the probability they will be able to ﬁnd a
job at or above the higher minimum wage.

For some teenagers, the disemployment effects of a minimum wage increase can
be offset somewhat by returning to school and raising the student’s skill level. To the
extent this occurs and is ignored in the analysis of the effects of minimum wages, the
negative impact of minimum wages on employment overstates the overall cost. Mattila
(1981) presents time-series evidence to support the hypothesis that teens return to school
when facing unemployment due to increasing minimum wages. The results indicate that
increases in the minimum wage move teenagers away from full-time work. Teens
replace the lost work time with additional schooling, perhaps with part-time work.
Mattila also addresses the indirect effect that additional teens present to schools. The
additional students returning to school increase the average class size, thus lowering the
quality of the education received by the median student. Also, unemployed students
forego on-the-job training opportunities, instead receiving additional education. This
may be a second-best approach to human-capital investment for many teenagers who may
learn best in a non-academic, skill-oriented environment.

Mattila’s results from time-series data were not supported in cross-sectional
studies (for example, Ehrenberg and Marcus (1980) and Neumark and Wascher (1996)).
Joint school and work decisions were found to involve higher-skill students taking jobs
away from lower-skill workers, as employers sought to increase the productivity of their
workforce in response to higher minimum wages. The cross-sectional studies also found

some evidence that workers displaced due to increases in the minimum wage are unlikely

126

to return to school. Overall, more recent research has challenged the validity of Mattila’s
ﬁndings.

Ehrenberg and Marcus (1980) use data from the 1966 National Longitudinal
Survey (NLS) younger male sample and the 1970 Census to investigate the effects of
minimum wages on educational outcomes. The stylized fact in labor economics is that
higher minimum wages reduce the demand for part-time workers. Increases in the
minimum wage will cause employers to seek more productive, full-time workers and
fewer less productive, part-time workers. Ehrenberg and Marcus ﬁnd evidence consistent
with this, using the NLS data for white male teenagers. Teens from low-income, white
families shift from being both enrolled and employed (most likely part-time) to being
employed full-time. Teens from higher-income, white families shift away from work
toward full-time schooling. However, for nonwhite families, increases in the minimum
wage were found to increase the probability a nonwhite male from a low-income family
is employed full-time, while reducing the probability they are in school full-time.

Evidence from the 1970 Census supports a different conclusion. Increases in the
minimum wage were found to shift jobs away from children from poor families to
children from non-poor families. This would suggest that higher minimum wages may
displace less-skilled workers (from poor families), and replace them with more-skilled
workers (from non-poor families).

A number of studies have investigated whether minimum wages affect the
probability a teenager or young adult will be enrolled in school. Cunningham (1981)
compared evidence from the 1960 and 1970 Censuses. The results were presented

separately for blacks and whites. The study also looked at two age groups, teenagers

127

aged 16-19 and young adults between 20 and 24. Among whites, Cunningham found
evidence that higher minimum wages resulted in reduced school enrolhnents among
teenagers. This is a striking result, given that Cunningham also found that minimum
wages have signiﬁcant negative effects on the employment of teens. However, the result
would be supported by future research (Ehrenberg and Marcus (1982) and Neumark and
Wascher (1996)). The standard hypothesis had been that newly unemployed teens
returned to school. The evidence reported by Cunningham is contrary to that hypothesis.
The results were similar for males and females. There was no evidence that minimum
wages signiﬁcantly affected school attendance among young adults.

Cunningham’s results are different for blacks. The minimum wage reduces
signiﬁcantly the probability a black teenager will not be in school, i.e., black teenagers
are more likely to attend school when minimum wages are higher. Black males were
more likely to be employed in jobs not covered by the minimum wage. These ﬁndings
were consistent with newly unemployed black teenagers returning to school, while white
teenagers may queue for jobs at the higher minimum wage without returning to school.

Leighton and Mincer (1981) examined the effects of minimum wages on wage
growth, training, and tenure. It is to be expected that increases in the minimum wage will
lead employers to substitute either hi gher-quality labor or capital for lower-skilled
workers. As a result, the prevalence of employer-provided training would be expected to
decline. Higher minimum wages would also be associated with ﬂatter wage proﬁles over
a worker’s career, as the employer adjusts with reduced training. Productivity growth
also would be reduced. The predicted results are that less training is provided and that

wage growth is slower. In samples of white and black men drawn from the Panel Study

128

of Income Dynamics and the National Longitudinal Survey of Youth, Leighton and
Mincer ﬁnd both of these effects. The impacts of minimum wages on wage growth and
training are negative and signiﬁcant for men with lower educational attainment (less than
high school completion). This is the group most likely to be affected. This is consistent
with higher minimum wages reducing the incentive for employers to provide training,
thus reducing future productivity growth.

Hashimoto (1982) uses data from the National Longitudinal Survey for 1966 and
1969. These years were chosen to bracket the minimum wage increase and coverage
expansion contained in the amendments to the Fair Labor Standards Act in 1967. One
way employers could adjust to higher minimum wages is to reduce training. Employees
who receive less training will have their productivity increase at a slower rate, resulting
in slower increases in wages. Hashimoto uses theoretical and empirical models to
evaluate the effect of higher minimum wages on the amount of training received by
employees. The theoretical model used by Hashimoto predicts that competition among
workers for jobs at the higher minimum wage will result in reduced employer
expenditures on training.

The empirical model was developed by Lazear (1979) and was adapted by
Hashimoto to test whether wage growth slowed after an increase in minimum wages.
Hashimoto ﬁnds evidence of reduced training after the increase in the minimum wage
that occurred in 1967. The results imply that a lO-percent increase would lower the
increase in wages due to additional experience by approximately 28 percent.

Ehrenberg and Marcus (1982) completed a follow-up study of the effects of

minimum wages and enrollment, using more sophisticated statistical methods. The

129

results again differ by race. For whites, teens from low-income families are more likely
to leave school for work at higher minimum wages, while teens from higher-income
families are more likely to increase their schooling. It is hypothesized that the teens from
higher-income families become more committed to school, both because part-time work
becomes more scarce with a higher minimum wage, and because the opportunity cost of
not being enrolled (a reduction in skill level that makes ﬁnding a future job more
difﬁcult) is increased. However, black male teens are seen to respond to a minimum-
wage increase by shifting from full-time schooling to full-time employment.

Neumark and Wascher (1996) used observations from matched CPS surveys to
investigate the response of young workers to minimum wage increases. Their results
indicate that lower-skill workers are displaced from their jobs as employers hire hi gher-
skill workers. Some of these higher-skill workers leave school to pursue jobs at the
higher wage. The net result is a reduction in teenage enrollment, and the lowest-skill
teenagers lose their jobs. These low-skill workers tend to be minorities and younger
workers (16 and 17 year-olds) who are displaced and potentially become unemployed and
non-enrolled. The result is an increased probability these teenagers become unemployed
long-term.

Neumark and Wascher (2001) attempt to examine the effects of minimum wages
on an individual’s efforts to acquire additional training on the job, or to qualify for a job.
The data they use include questions on training included in the January 1983 and 1991
Current Population Surveys. Using interstate variation in the minimum wage rate to
identify the impacts of minimum wages on training, they ﬁnd evidence that higher

minimum wages reduce on-the-job training. This is consistent with the ﬁndings of

130

Leighton and Mincer and Hashimoto. Their results imply that formal training is more
adversely affected by higher minimum wages. They argue this is to be expected if formal
training entails more direct and indirect costs. Informal training is more of a leam-by-
doing process, and is thus a joint product of the production process.

Neumark and Wascher also investigate whether higher minimum wages result in
increased training in order to obtain a job. Their results indicate there is no such effect.
While minimum wages reduce on-the-job training, they do not seem to encourage
individuals to acquire additional skills prior to obtaining a job. This would seem to imply
that minimum wages reduce the skill-level of workers overall, thus lowering future wage
payments. Neumark and Wascher note that their data may not provide complete enough
information on very general skill acquisition.

The studies discussed above are summarized in Table 6.3. The evidence on the
effects of minimum-wage increases on different demographic groups is mixed. Some of
the earlier studies, Mattila for example, indicated that teens returned to school following
increases in the minimum wage. This is consistent with teens being displaced from jobs
when the minimum wage increases, and those displaced teenagers returning to school in
order to increase their productivity.

Later studies by Ehrenberg and Marcus (1980 and 1982) found that higher
minimum wages result in an employment/schooling shift. Teens from low-income
families tend to move out of school and into full-time work when minimum wages rise,
while teens from higher-income families leave the labor market and focus on school.
There is also evidence that employers reduce training when minimum wages increase

(Leighton and Mincer (1981), Hashimoto (1982), and Neumark and Wascher (2001)).

131

Taken together, these ﬁndings imply that minimum wages draw teens from low-income
families away from school, and into jobs that feature less training and slower wage
growth. Alternatively, teens from hi gher-income families return to school to receive
additional training.

The adverse effect of the minimum wage may be even higher in light of the
ﬁndings of Neumark and Wascher (1996). They report that higher minimum wages
result in an employment shift from low-skill teens to higher-skilled teens. They do not
ﬁnd evidence that the newly unemployed, low-skill teens return to school.

When the results are combined, they suggest that higher minimum wages lower
the skill accumulation of teens currently possessing the lowest skill level in two ways.
First, these teens leave school sooner to take jobs with relatively low levels of training.
Second, higher minimum wages make them vulnerable to layoff and replacement,
resulting in unemployment. This observed outcome would seem to demonstrate that the
minimum wage is not a precise tool for improving the long-term prospects of the lowest-

skilled teens.

E] T C and Capital Accumulation

Eissa and Hoynes (1998) investigated the labor-market effects of the EITC on
married couples. From the perspective of this study, the most relevant ﬁnding they
present is that expansions of the EITC are associated with declines in participation and
hours worked by married women. This may be due to the high marginal tax rates faced
by families in the phase-out range of the EITC. If the families in this study are in the

phase-out range, the results are consistent with the economic theory discussed in Chapter

132

2. Since the sample used by Eissa and Hoynes only contains couples where the wife has
less than a high-school education, it is possible that the reduction in hours could be a
result of the wife pursuing additional schooling. The authors do not address the question
of how the time no longer spent in the labor market is used. It is possible the additional
time is spent in parental activities or in other household production, as well as potentially
attending school or vocational training.

The second body of literature concerns how EITC recipients spend the credit they
receive. The focus is whether the credit ﬁnances increased consumption or capital
investment. The interest in the potential impact of the EITC on capital accumulation
comes from the fact that the EITC is at least partially available in regular installments or
entirely as a lump sum at the end of the tax year. Almost all recipients receive the credit
as a lump sum (Barrow and McGranahan, 2000). This is consistent with the possibility
that the EITC affords low-income households the opportunity to “save” in order to
ﬁnance capital investments. These investments might take the form of improved
housing, transportation, or education. However, the lump-sum delivery system may also
indicate that recipients are unaware of the advance-payment option, or that signiﬁcant
bureaucratic barriers exist to receiving the EITC in this manner.

In general, studies on the use of the EITC consist of surveys of credit recipients.
The reported results seem to indicate that the recipients spend a signiﬁcant portion of the
reﬁrnded credit to ﬁnance current consumption, pay down or catch up on debts, and to
ﬁnance lump-sum capital expenditures. There are relatively few papers on this subject.

Smeeding et al. (2000) surveyed clients of a free income-tax preparation service

provided by the Center for Law and Human Services in the Chicago area. The clients

133

used for the study were both receiving a federal tax refund and claimed the EITC. The
survey, among other things, asked recipients about their intended use of the tax refund
and EITC. The survey respondents were asked to list and prioritize up to three uses of
their refund. The responses were then classiﬁed as either expenditures to “make ends
meet” or to provide upward social mobility. The ﬁrst category includes spending on
regular bills (e. g., food, rent, utilities), clothing, appliances, and furniture. The second
category includes debt repayment, savings, and capital investments for schooling,
transportation, and housing. Most of 650 respondents provided at least one spending
priority (81 percent) with fewer providing additional responses (46 percent with at least
two and 14 percent with three).

The most commonly cited spending plan for the tax refund was paying bills (50
percent listed as their ﬁrst priority) with current consumption purchases (food, clothing,
etc.) second. More than one-half of all respondents indicated they would use the refund
to improve their upward mobility. While the expenditure category is deﬁned very
broadly, this is consistent with the EITC having longer-term beneﬁts. More to the point
of this study, 16 percent of respondents listed tuition payments among their top three
choices. This could be tuition for the respondent, spouse, or child. This compares with
22 percent who listed an automotive-related expense among their top three choices.

The study is difﬁcult to generalize because the sample of taxpayers responding to
the survey is nonrandom. As the authors point out, the free tax service did not offer
“rapid refund”, while many professional tax services did. It is likely that EITC recipients
most in need of their refund would choose a ﬁrm like H&R Block, as opposed to using

the free service. This would bias the results toward greater social mobility investments.

134

It may also be that gathering data from one city in the upper Midwest produces results
that are not generally true around the nation.

Romich and Weisner (2000) survey 42 households participating in the New Hope
Project in Milwaukee, Wisconsin. This is a community-based program designed to
alleviate poverty. The program involves more than 1,300 households, from which a
random sample of 60 households was drawn. The study is based on the households
agreeing to participate. The results are based on surveys of the remaining households.
The two questions addressed by this study are:

1. Which delivery mechanism gives greater utility, the advance payment option or
the lump sum?

2. What kinds of purchases will be made with the respective payment forms?

The sample selection and size make it difﬁcult to generalize the results, but this
study provides further evidence that families receiving the EITC use the lump-sum
payment of the credit to allow them to purchase consumer durables, invest in housing,
and pay tuition. The tuition expenses referred to in this study are focused on paying
private school tuition for children, not additional schooling for adults. Even if the EIT C
is associated with reduced enrollments by parents, the credit may allow for increased
investments in schooling for the children of the recipients.

Barrow and McGranahan (2000) use data from the Consumer Expenditure Survey
to examine the spending patterns of households eligible for the EIT C. The authors
compare the seasonal spending patterns of households eligible for the EITC with the
spending patterns of households that are not eligible. Eligible households show some

evidence of increased expenditures in the months of likely receipt of the lump-sum tax

135

refund (including the EITC), especially for durable goods. The study also ﬁnds that
spending increases by less than the average EITC. This would indicate that households
save part of their credit, at least temporarily, or engage in consumption smoothing. This
study does not directly relate to the task here, but given the ﬁnding that households do
save some of their credit, the study is consistent with EITC recipients using their credit to
make human-capital investments.

The results from the EITC literature are summarized in Table 6.4. There is some
evidence that some recipients of the EITC may use the EITC for human-capital

investments. However, the evidence is not strong.

6.4. MODEL

This study will look at two issues. First, evidence will be presented that is
consistent with earlier studies on the effects of the EITC on the labor supply of single
mothers. There are a number of studies, including Eissa and Liebman (1996), Meyer and
Rosenbaum (2001), and Ellwood (2000), that conclude the expansions of the EITC have
led to increased labor-market participation among single mothers. These studies were
summarized in Chapter 2.

The purpose for re-examining the topic here is to investigate the EITC’s
effectiveness in increasing labor supply among younger women. Younger women with
children may have family support that allows them to postpone labor-market activities
and complete school, especially high school. This study will focus on women between
the ages of 17 and 21. These are prime school years where women may complete high

school and attend an initial post-secondary program, which could be either vocational or

136

academic. Many women also either postpone or skip post-secondary education, instead
pursuing employment.

Having a child during these years may affect the opportunities available to young
women. A very young child requires constant care. A single parent would need to ﬁnd a
child-care provider in order to continue education or work. If a family member cannot
provide these services, they will generally be purchased at a signiﬁcant cost.

In the past, social welfare programs, such as Aid to Families with Dependent
Children (AFDC), have provided beneﬁts to single mothers so that they could care for
their children.1 In recent years, support for such programs seems to have eroded. It has
been replaced with interest in supporting single parents in the labor market, among other
things.2 Speciﬁcally, expansions of the earned income credit and the increased use of
welfare waivers allowed families to earn higher amounts of labor income without losing
public assistance beneﬁts. This perhaps culminated in the passage of welfare reform, the
Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA).
It seems clear that the policy focus has shifted from solely providing a safety net to
encouraging and, in fact, requiring that single parents work to provide their own support.3

As mentioned above, the ﬁrst part of this study will focus brieﬂy on the effects of
the expansions of the EITC on the labor supply of young women. The goal is to be able

to compare the impact of the EITC expansions on this small subset of all single women

with the effects on larger samples reported in the literature. If the EITC has affected the

 

' Beneﬁts were rapidly phased out with increased labor income, thus making AFDC primarily an income
support program.

2 There has also been an increased focus on maintaining two-parent farrrilies, one of the ofﬁcial goals of
PRWORA.

3 Schooling and training programs, while perhaps not as openly discussed as work requirements, often
satisfy the work requirements of welfare reform plans.

137

IE

labor supply of young women in a manner similar to the credit’s effects on the larger
population of single mothers, it might be expected that increased work activity would
come about because of reduced schooling.

It is not necessary that single mothers who increase their labor supply reduce their
school attendance. It is possible that increases in the labor supply of young mothers over
time have resulted in reduced proportions of young mothers who are neither employed
nor enrolled in school. It is also possible that the expansions of the EITC have made it
possible for single mothers to both work and attend school. Once the effects on labor
supply are estimated, the study will turn to the effects on school enrollment.

In order to estimate the effects of the EITC on labor supply, a probit model is

estimated. The model can be expressed as:

P(po,~t = 1) = <1>(oz + ﬁlm + 0Y2 + yokids + 71(EITC control) +

y2(state EITC control)) + 8,,

where is P(poit) represents the probability that an individual is employed during the year.
Z,, is a vector of individual demographic variables such as age, educational attainment,
race, number of children under 6, and non-labor income. Y“ is a vector of state control
variables, including the unemployment rate, the maximum AFDC beneﬁt for a family of
three, the growth rate of personal income, the ratio of the minimum wage to the average
state hourly wage, and adjustments to welfare policy as reﬂected by the presence of a
statewide welfare waiver. e is a random error term that is assumed to have a standard

normal distribution.

138

The EITC control is the wage subsidy rate each woman would face upon entering
the labor market. This is equal to zero if the woman has no children, and up to 40 percent
if the woman has two or more children in 1996. The variable is intended to model the
policy change over time, and does not reﬂect the actual subsidy or tax rate faced by
individual observations. Two speciﬁcations of the state EITC variable are used. The ﬁrst
is the wage subsidy rate upon participation for all states with EITCs. This variable is
only nonzero for women with children living in a state with an EITC. The second
speciﬁcation limits the state EITC to the wage-subsidy rate upon participation for states
with refundable EITCs. 8 is a random error term that is assumed to have a standard
normal distribution.

This equation is estimated for a random sample of single women, drawn from the
March Current Population Survey for the years 1987 through 1996. The women were
between the ages of 17 and 21 when surveyed. This would place the women in an age
group where school attendance includes completing high school and beginning post-
secondary education, and early in the range of childbearing ages. The estimates are
repeated while expanding the sample to include single women between the ages of 17 and
24.

The focus of the study then shifts to estimating the effects of the EITC on school
enrollment. With school enrollment added to the study, a woman may now be in one of
four possible activity categories. Each woman has an enrollment status and an
employment status, as reﬂected by her answers to the March CPS questionnaire.

This results in four possible outcomes: not in school/not employed (N SNE), not

in school/employed (NSE), in school/employed (SE), and in school/not employed (SNE).

139

Consistent with the Ehrenberg and Marcus (1982) and Neumark and Wascher (1996)
studies, a multinomial logit model is used to estimate the effects of the EITC on

school/employment status.

The utility from each activity a in state 5 and period t for individual i is

Uiast = or + [Slim + 7Y3. + (1) (Federal EITC) + 6 (State EITC controls) + sis, .

As described above, list represents a vector of demographic variables and Y,,
represents a vector of state variables designed to control for differences in policy regimes
and economic environments across states. Estimates were obtained using state dummies
to control for long-term, ﬁxed effects. However, since the results including state
dummies are not signiﬁcantly different, those results are not presented here.

The federal EITC variable is the phase-in rate of the EITC for each woman,
depending on the year and the number of children she has. The subsidy rate has
increased over time, beginning at 10 percent in 1975 and reaching 40 percent in 1996.

State-level EITC controls were again included.

6.5. DATA

The model uses a sample of women between the ages of 17 and 21 drawn from
the March Current Population Survey (CPS). The sample covers the period from 1987
through 1996, just following the enactment of the ﬁrst expansion of the EITC (with the

Tax Reform Act of 1986) and concluding as welfare reform is enacted (PRWORA 1996).

140

The dependent variables were constructed using questions from the March CPS on school
enrollment and employment status in the previous year. A young woman was considered
enrolled if she reported being either a full- or part-time student. A young woman was
considered employed if she reported working at least one week in the prior year.

This young sample is used to focus attention on the potential impact of the EITC
on women with a greater probability of being enrolled in school, and for whom having a
child is an unusual event. With three-fourths of women 24 and under without children,
the women in this sample with children are signiﬁcantly ahead of the average of their
peers in terms of fertility.4 The regressions were repeated using unmarried women
between 17 and 24 without signiﬁcant qualitative differences.

Demographic variables were included for age, race, education level, unearned
income, a dummy variable indicating whether a women had at least one child, and the
number of children under 6, all of which came from the CPS. An interaction term was
included for race and the child dummy. Variables were used to control for variation
across states in economic conditions and environment. These included the state’s
unemployment rate, the growth in real per-capita personal income, the relative value of
the minimum wage, the value of the maximum grant for a family of three from the Aid to
Families with Dependent Children program (AFDC), and a variable that indicates the
number of months during the year that the state of residence had a statewide welfare
waiver in place. An interaction term between waivers and the child dummy was also
included. Finally, a time trend was included to capture long-term effects.

Data on unemployment rates were obtained from published data from the Bureau

of Labor Statistics, U.S. Department of Labor. Real per-capita personal income data are

141

from the Bureau of Economic Analysis, U.S. Department of Commerce. The data on
state-level minimum wages are from the U.S. Department of Labor. Data on the AFDC
program are from the Green Book published periodically by the U.S. House of
Representatives Committee on Ways and Means. The data on welfare waivers are taken
from Council of Economic Advisors (1997).

The federal EITC was modeled using a variable that equaled the maximum wage-
subsidy rate for which the woman would have been eligible under the tax credit. For
women with no children, the subsidy rate is zero. For women with at a least one child,
the variable would equal 14 percent for 1987 through 1989. The two expansions of the
federal EITC during the 19905 increased the subsidy rate and created higher subsidy rates
for women with two or more children. Controls were also introduced for state-level
EITCs, so that interstate variation in the application of these credits would provide
additional identiﬁcation of the credit’s effectiveness. The state-level EITC controls were
the wage-subsidy rate for each state with an EITC, and a second speciﬁcation limiting
this control to only states with refundable EITCs.

Summary statistics for the sample are presented in Table 6.5. The sample
includes data on 43,533 single women between ages 17 and 21. As expected, the average
woman in the sample is approximately 19 years old, and has attended some college.
Over the course of the sample period, less than 10 percent of the sample reported having
had a child. For white women, 6.1 percent had at least one child. For nonwhite women,
20.3 percent had at least one child. Approximately 68 percent of the young women

surveyed were currently enrolled in school, while 71 percent were currently employed.

 

4 U.S. Census Bureau, Table 1, Women and Fertility 2000.

142

Approximately 20 percent of the women in the sample are nonwhite. Of the
women in the survey, 47 percent report being both employed and enrolled during the past
year. Almost equal proportions, 24 and 21 percent, report working only and school only,
respectively. Only 7.7 percent reported being both non-enrolled and non-employed in the
past year. Sample statistics are also presented in Table 6.5 for single women between the

ages of 17 and 24.

6.6. PROBIT RESULTS ON PARTICIPATION

Estimates are presented for labor-force participation, in order to demonstrate the
broad consistency of the sample used for this study with the results from prior studies.
Prior studies have focused on labor-force participation as an indicator of the success of
the EITC. Estimates of the effect of the EITC on participation are presented here in order
to present additional evidence supporting those results using a slightly different policy
framework, and to then expand the investigation to the impact the EITC may have on
school enrolhnent.

The ﬁrst set of probit estimates is presented in Table 6.6, and includes the results
for both age groupings. The explanatory variables are age, a dummy variable indicating
the presence of a child, number of children under 6, unearned income, years of education,
and race. An interaction term is added for race and children. A trend variable was also
included. Each equation includes the EITC wage-subsidy rate applicable to each woman.
Women without children would not be eligible for the EITC, and would thus have a

subsidy rate equal to 0.

143

The estimates imply a 7.6 percentage-point increase in labor-market participation
for an additional year of age. Women closer to 21 are more likely to be employed. This
is to be expected, because women complete high school as they pass 18 years old, and a
signiﬁcant percentage do not continue in school beyond high school. The probability of
being employed is signiﬁcantly higher for the women in the sampler who are closer to 21.

The presence of an additional child under 6 reduces labor-market participation by
almost 16 percentage points. An additional year of education increases the probability of
working by approximately 1.2 percentage points. Women who are not white are less
likely to work than white women. The estimates here imply that nonwhites are 22
percentage points less likely to work than white women, all else equal.

The results presented in Table 6.6 are consistent for both age groupings. The
sign, general magnitude, and statistical signiﬁcance of the coefﬁcient estimates are
practically the same for both groups. The small variations that occur in the estimates
result in estimates that are generally smaller in absolute value for the sample with women
ages 17-24. The one exception is for the presence of a child. The results indicate that
many factors affect the labor supply of women ages 17-21 to a greater extent than the
impact on women just slightly older.

The positive coefﬁcient estimate for the EITC implies that women with children
have a greater probability of participating in the labor market later in the sample period
than in the early years. The EITC variable increases over time, to reﬂect the law changes
that raised the wage subsidy rate during the 19905. The estimated partial derivative is
0.748 for women ages 17-21, and 0.674 for women ages 17-24. These coefﬁcient

estimates are very large, and imply an increase in labor-force participation of between

144

11.1 and 12.3 percentage points among young women with children based on the 1993
EITC expansion for women with one qualifying child. Applying the estimated partial
derivative of 0.884 to the 1990 EITC expansion, which was much smaller in magnitude,
would imply an increase in participation of approximately 2.5 percentage points.

The large coefﬁcient estimates for the EITC may help explain why, in Chapter 3,
the estimated effect of the EITC was smaller among older women (ages 25-50) than some
estimates in the existing literature obtained with samples that included women between
ages 18 and 24. From the estimates obtained here, it appears that the difference is
attributable to the large effect on the participation of younger women. The results here
imply that the EITC has a strong effect on participation among the youngest single
mothers. The estimated increase in participation presented by Meyer and Rosenbaum
(2000) was 8.5 percentage points for women ages 19-44. The estimate here is larger,
consistent with younger women being more responsive to policies designed to encourage
work.

The regressions above were repeated with a set of state variables that are designed
to control for interstate variation in economic conditions and policy environments,
including two speciﬁcations with state EITCs. The results are presented in Tables 6.7
(ages 17-21) and 6.8 (ages 17-24). The results are consistent with those presented in
Table 6.6. All of the estimated coefﬁcients presented above retain the same estimated
sign, remain strongly signiﬁcant, and are of similar magnitudes. This is true for both age
groupings. There is some small variance in the estimated effects of a child and additional

education, but the magnitude of these changes is very small.

145

Some signiﬁcant effects are observed among the new state-level controls. Higher
minimum wages are estimated to result in higher labor-force participation among this
sample of young, unmarried women. The estimates are strongly signiﬁcant, and imply
that a 10-percentage-point increase in the relative minimum wage would increase
participation among single women by approximately 4.5 percentage points. This is not
consistent with theoretical predictions that higher minimum wages affect younger
workers adversely. This result is also different from the results presented in Chapters 3
and 4. Among older women, minimum wages were found to have an insigniﬁcant effect
on labor-market participation. However, some empirical evidence does exist to indicate
that minimum-wage increases have differential effects across demographic groups.5

Welfare policy is generally estimated to have a negative effect on labor-market
participation. Higher welfare beneﬁts are correlated with reduced participation.
However, the estimate is signiﬁcantly different from zero for the sample with women
ages 17-24 only. The estimate for this group implies a $100 increase in monthly beneﬁts
would reduce labor-market participation by 1.9 percentage points. The estimate for
single women ages 17-21 is very small and negative. The estimates for the variable
controlling for waivers are negative, the opposite of the expected sign. Neither estimate
is signiﬁcant at traditional levels. These ﬁndings together indicate that welfare policy has
not played a statistically signiﬁcant role in the increased labor-market participation of
young single mothers.

Economic conditions are estimated to have a signiﬁcant effect on the labor-
market participation of younger women. As was found in Chapter 3, higher

unemployment rates within a state reduce employment among single women. However,

 

5 For a summary of the research, see Brown, Gilroy, and Kohen (1982).

146

the effect is much larger on this younger population. The estimates here imply a 1.2
percentage-point decline in employment for each one-percentage-point increase in the
unemployment rate for women between ages 17 and 21. The effect is smaller for the
slightly older sample.

Higher per-capita income growth is also associated with lower employment for
single women. The estimated effect of a one percentage-point increase in income grth
is a 1.1-percentage-point reduction in employment of single women. While this result
may not agree with initial expectations, there may be a possible explanation. It may be
that single women are less likely to work because they receive greater economic support
when economic times are good. Especially for this sample of younger women, greater
economic opportunities for members of each woman’s support network (e. g., family or
household members, boyfriends) could allow more women to participate in non-market
activities, such as school or care for a home.

For single women between the ages of 17 and 21, the estimated effects of the
federal EITC remain positive and signiﬁcant, with an estimated partial derivative
approaching 0.9. This estimated effect is very large. Based on this estimate, the 1993
expansion of the EITC on average increased the participation of single mothers between
the ages of 17 and 21 by 14.8 percentage points. The estimated effect is smaller when the
sample is expanded to include all single women age 17 to 24. For the larger sample, the
estimate is 0.75. This is still a very large estimate, consistent with the conclusion that the
federal EITC has signiﬁcantly increased labor supply among young single mothers.

The effect of a state EITC is estimated to be negative overall. In both sets of

estimates, the impact on participation for all state EITCs (refundable and nonrefundable)

147

is negative and signiﬁcant. The estimates are large in magnitude, with an estimated
coefﬁcient of —1 .8 for women ages 17 to 21, -1.2 for women ages 17 to 24. There is
some change when the control variable for state EITCs is restricted to refundable credits.
A reﬁrndable credit should have a larger positive effect on labor supply because any
excess of the credit above state tax liability is refunded. When only refundable state
EITCs are used, the estimates become smaller in absolute value and insigniﬁcant. This is
true for both age groupings.

These estimates imply that state EITCs are not particularly effective at quickly
moving single mothers into the work force. This is consistent with most of the estimates
in Chapter 3. As discussed there, it is likely that this estimate is inﬂuenced strongly by
New York’s EITC that was adopted in 1994. Given New York’s relative size and
economic situation in 1994 (also discussed in Chapter 3), this estimate may not reﬂect the
results in other states.

However, this is not found to be the case for younger women. When the
observations from New York are removed from the sample, the results are basically
identical. While the results are not presented in this paper, the estimated coefﬁcients for
state EITCs remain unchanged, both qualitatively and quantitatively. The evidence here
indicates that state EITCs do not, on their own, increase labor supply among single
mothers under age 25, with other factors unchanged.

The results presented here are consistent with the growing body of economic
research on EITCs discussed more fully in the Chapter 2. The expansions of the federal
EITC are associated with increased employment among single mothers, and these effects

are quite strong among this younger sample. The results are not as positive for state

148

EITCs, although the estimated effect of reﬁrndable credits is statistically insigniﬁcant.
Economic conditions such as higher unemployment result in large decreases in
employment. Welfare policy changes such as waivers are not found to have signiﬁcant

effects in increasing the employment of the single mothers in this sample.

6.7. MULTINOMIAL LOGITS ON ENROLLMENT STATUS

This paper now turns to the core topic to be investigated, whether the expansions
of the EITC, which have consistently been estimated to increase labor-market
participation among single mothers, have resulted in a shift away from schooling. In
order to investigate this question, a multinomial logit model is used. For all of the
estimates presented here, the base case was assumed to be non-employed and not
attending school (N SNE).

The results are presented in Tables 6.9 through 6.14 for women ages 17 to 21.
The results for single women ages 17 to 24 are presented in Tables 6.17 through 6.22.
Three speciﬁcations are estimated. The ﬁrst set of estimates was obtained in a manner
similar to the results presented in Table 6.6 for labor-market participation. The set of
explanatory variables includes only basic demographic variables and the federal EITC
variable. The estimates are presented in Table 6.9, with the implied partial derivatives
presented in Table 6.10. The partial derivatives were evaluated at the sample means for a
woman eligible to receive the EITC.6 The estimated coefﬁcients presented in Table 6.9

are all signiﬁcantly different from zero at traditional signiﬁcance levels.

 

6 The partial derivative was calculated using values of 1 for children (dummy), 1 for child under 6, and a
value of 0.34 for the EITC subsidy rate.

149

Additional years of age imply reductions in school attendance. An additional year
of age is estimated to result in approximately an 11.8-percentage-point decrease in the
probability of being in school. This is predominantly due to an increase in women taking
jobs (increases 13.2 percentage points) and a reduction in the probability a woman is
attending school exclusively (declines 8.1 percentage points).

As would be expected, women with children have a lower probability of both
working and attending school, as reﬂected by the partial derivatives reported in the third
column of Table 6.10. Having at least one child is estimated to reduce the probability of
both working and attending school by more than 30 percentage points. There is a strong
increase in the probability of being simultaneously out of school and the workforce when
a single woman has a child (21 percentage points).

Employment is less likely for women who are not white. Nonwhites are more
likely to be attending school but not employed, or neither attending school nor employed,
than are white women. For nonwhites with children, the results are different. While
having children or being nonwhite are estimated to reduce the probability of
simultaneously attending school while working, for nonwhites the probability of
participating in both activities increases with the presence of a child. This increase is
offset by a decrease in the probability of being non-employed and non-enrolled. This
result is difﬁcult to explain, but the estimates are statistically signiﬁcant, and robust
across all speciﬁcations and both age groupings. There may be a renewed focus on
economic issues once a nonwhite woman becomes a mother, although it is unclear why
any change in focus would affect various races differently. This result may reﬂect

differences in the support network available to the average single mother across races.

150

The federal EITC is estimated to have a signiﬁcant effect on increasing the
probability a single mother will be employed, both with and without attending school.
The estimates in Tables 6.9 and 6.10 imply that a large part of the increased work/school
activity among single mothers is associated with a decreased probability that a single
mother is unemployed and not attending school. The estimated partial derivative implies
that a 10-percentage-point increase in the wage subsidy rate would lead to approximately
a 2.0-percentage-point increase in the probability a single mother both attends school and
works. Similarly, the probability that a woman eligible for the EITC works, but does not
attend school, increases by between 2.9 and 3.5 percentage points with a similar 10-
percentage-point increase in the wage subsidy rate.

Similar results are obtained for single women ages 17 to 24. For women with
children, the federal EITC is estimated to have increased the probability that a single
woman is employed, both with and without attending school. The estimates are
statistically signiﬁcant. The largest increase is in the probability that a single mother
both works and attends school, although there is not a meaningful difference between this
category and working without going to school. The increased work activity comes from
a reduced probability that single mothers are both non-employed and non-enrolled. The
estimated impact on being enrolled only is not statistically signiﬁcant.

Given the results from previous research as well as those presented above, it is not
surprising that the EITC results in increased labor supply while out of school. However,
the estimate that the EITC is associated with a statistically signiﬁcant increase in the
probability a young mother is employed an_d enrolled is somewhat surprising. Three

possible factors could be contributing to this effect. First, increases in the wage-subsidy

151

rate would provide single mothers, who often work at lower-wage jobs, with a signiﬁcant
potential increase in the resources available to them, assuming there is no signiﬁcant
decrease in the number of hours worked. These additional resources would potentially
provide opportunities for the woman to afford to pay for schooling, or for additional
childcare so that the mother could attend classes.

A second explanation is that the expansions of the EITC have raised the return to
work, so that single mothers now ﬁnd it less desirable to be out of the work force while
attending school. The estimated effect of the EITC on being enrolled but non-employed
presented here for both age groups is not statistically signiﬁcant in any speciﬁcation. The
coefﬁcient estimate is more often positive, indicating some weak correlation between
increased school attendance without employment for some single mothers. This may
reﬂect increased opportunities available to single mothers who wish to continue their
education, perhaps due to improved social infrastructure (e. g., increased access to
daycare).

A third explanation is that if the EITC results in an increase in lifetime labor
supply, single mothers may also choose to seek out more schooling in order to make the
additional labor supply more proﬁtable. If a worker will spend additional years in the
work force, this will provide additional time to recover the cost of investments in
education. Also, additional education is more valuable when the length of time the
individual will spend working increases. Thus the wage subsidy available to single
mothers through the EITC may simultaneously encourage entering the labor force and

returning to school.

152

Tables 6.11 through 6.14 present the coefﬁcient estimates and partial derivatives
when the model is estimated using state-speciﬁc controls for economic conditions and
policy variables. For most of the variables discussed above, the estimates do not change
signiﬁcantly. As such, the estimates will not be discussed again. State dummies were
also included to control for state-level ﬁxed effects. Once again, the results for the
estimates of interest do not change materially, and the results are not presented here.

The addition of state controls does not result in signiﬁcant changes in the
estimated probabilities for the federal EITC variable with regard to employment. A
higher wage-subsidy rate under the federal EITC is estimated to increase the probability
that a single mother is employed, and reduce the probability she is non-employed and
non-enrolled (N SNE). This is a consistent ﬁnding with both sets of economic controls,
and is repeated in the estimates for single women ages 17 to 24 (Tables 6.17 through
6.22). This result is also consistent with other studies of the EITC that ﬁnd the credit
leads to increased labor-market participation. The estimates imply that expansions of the
federal EITC have increased enrollments as well, although the partial derivative for the
EITC is larger (0.35) for SE than for NSE (0.20). This implies the increase in labor
market participation due to the EITC has not resulted in decreased school attendance, but
rather increased the probability a single mother would be enrolled. The EITC estimates
obtained using the sample of single women ages 17-24 are slightly larger for SE than for
NSE, implying that the probability that single mothers have moved into both additional
schooling and work is slightly higher than the probability they have moved into work
alone. This result ﬁts the hypothesis stated earlier that the additional disposable income

provided by the EITC could allow for both increased schooling and increased labor

153

supply. It seems that the EITC provides single mothers with increased options to
improve their economic position.

However, state EITCs are not associated with statistically signiﬁcant increases in
the probability of employment. Even refundable state credits are not found to lead to
increased employment. The estimate that is most signiﬁcant statistically is a negative
effect on the probability of being employed and enrolled simultaneously.

The apparent disconnect between state and federal EITC policy is not intuitive. It
seems unusual to ﬁnd different results for the federal and state EITCs. It may be that
recipients understand federal policy better than they understand state policy. Whatever
the reason, the difference between the measured effectiveness of the federal and state

EITCs is striking.

6.8. ALTERNATIVE CALCULATIONS

A second method of evaluating the estimates from a multinomial logit model is to
present changes in probabilities for each activity, based on changes in the explanatory
variables (see Wooldridge (2002), pp. 497-499). This method will produce an estimate of
the effect of changes in EITC policies based on the change in the likelihood function.
The probabilities are calculated using the coefﬁcient estimates presented in Table 6.11.
The probabilities are calculated by evaluating the likelihood function using the
coefﬁcient estimates and a set of explanatory variables. Examples are presented in
Tables 6.15 and 6.16.

These estimates are going to be similar to the partial derivatives, but provide a

simple alternative calculation of the effect of changing the relevant policy variable. For

154

this analysis, the policy variable is the federal EITC. The effect is measured using the
coefﬁcient estimates from the multinomial logit estimation process, along with values for
all the explanatory variables. For most explanatory variables, the sample mean is used.
These values are then used to generate a probability of each particular
employment/enrollment status.

The ﬁrst example (Case 1 in Table 6.15) is for an increase in the EITC wage
subsidy rate from 40 to 45 percent. The example is based on a 17-year old woman with
two children under six. This hypothetical woman also has completed the 10th grade. The
woman is white and has approximately $200 in annual unearned income. The economic
variables are evaluated at their respective sample means. The coefﬁcient estimates imply
that the 5-percentage-point increase in the EITC subsidy rate would result in a decline in
inactive women (N SNE) of 2.5 percentage points, with an increase in the probability of
employment of 2.3 percentage points, comprised of a 1.4 percentage-point increase in no
school but employed (N SE) and a 0.9 percentage-point increase in school and work (SE).
These imply partial derivatives of 0.28 and 0.18 for the EITC variable for NSE and SE,
respectively. These results are generally consistent with those presented in Tables 6.12
and 6.14, although the estimated partial derivatives are smaller. However, the
implication remains clear. The EITC is estimated to move single mothers into
employment without a signiﬁcant decline in attending school. However, because this
analysis does not evaluate whether single mothers are attending full- or part-time, it is
impossible to know whether the amount of time spent in school increases or decreases.

The second example is based on the 1993 EITC expansion, and is Case 2 in Table

6.15. It compares the probability of being in each category in 1992 and 1996 for a

155

woman without children (and thus ineligible for the EITC) and a woman with two
children under 6. These hypothetical women were both 17, had ﬁnished 10th grade, had
$200 in unearned income, and were white. The results imply virtually no change for
women without children, while women with 2 children would be on average 10.0
percentage points more likely to work (obtained by summing the difference for NSE and
SE). The implied increase in work would be the result of a 6.4-percentage-point increase
in work without school (N SE) and a 3.6-percentage-point increase in work with school
(SE). This is also consistent with the EITC increasing employment among single
mothers without decreasing the probability a single mother attends school. There is also
an estimated increase of 2.4 percentage points in the probability that women attend
school without working.

The examples from Table 6.15 are repeated in Table 6.16, with the exception that
the single mother only has one child. As a result, the subsidy rate is lower, (0.34 in 1996,
as opposed to 0.4 for a mother with two or more children). For Case 1, the policy change
being evaluated is an increase in the wage-subsidy rate from 0.34 to 0.40. The results are
very similar to those in Table 6.15, perhaps not surprising given the small variation
between the subsidy rate, 0.34 versus 0.40. The probability that a single mother works
with the higher subsidy rate increases by 2.7 percentage points. This is slightly higher
among mothers with one child than the 2.3-percentage-point increase for mothers of two
children discussed above. The composition is different, though. Women with one child
have a greater probability of both working and attending school (SE), than women with
two children. That women with one child have a higher probability of both working and

attending school than women with two children is a reasonable result, considering that a

156

single mother of multiple young children would have increased parental responsibilities.
Also, multiple children would require increased childcare expenditures, potentially
limiting the resources available to purchase training.

Case 2 in Table 6.16 examines the effect of the 1993 EITC expansion on women
with one child compared with women without children. As in Table 6.15, there is a
signiﬁcant increase in the probability that a woman with one child works, either with or
without going to school, relative to women without children.

It is important to remember that these estimates are comparing the relative
probabilities. For example, the results in Case 2 of Tables 6.15 and 6.16 do not imply
that women with children are more likely to work than women without children after the
1993 expansion of the EITC, but rather that the relative probability increased. There was
an increase in the probability a woman with at least one child would be working, while

the probability for women without children remained unchanged.

6.9. COMPARISON WITH PRIOR LITERATURE

The estimates presented here are the ﬁrst evidence of the effects of the EITC on
the human-capital accumulation of single mothers. Because the credit presents a subsidy
to enter the labor market, it might be expected that single mothers do so at the expense of
additional schooling. However, the evidence here does not support that.

Instead, the higher wage-subsidy rates provided by the expansions of the EITC in
1990 and 1993 are associated with both increased labor supply and increased
participation in both school and work among single mothers. There is no evidence of

signiﬁcant declines in school enrollment.

157

The case is less clear for state EITCs. The estimated coefﬁcients for state EITC
variables are often insigniﬁcant. Reﬁlndable EITCs are associated with some increased
work while not enrolled, with a corresponding small decline in school enrollment. This
may represent more intensive effort, among the states with refundable EITCs, to
encourage work activity among single mothers.

One might wonder who these women are who respond to the expansions of the
EITC by not only increasing their labor supply, but also continuing their education? One
possibility is that the change in incentives brought about by the expansions of the EITC,
coupled with changing attitudes toward public assistance, may have resulted in
behavioral shifts among young women with stronger family connections. These
connections may include support with childcare as well as ﬁnancial assistance. Instead of
receiving public assistance and not working, it may be that these women have opted to
continue their education while also working part-time in order to claim the EITC.

The prior literature on minimum wages found mixed evidence on the effects that
higher minimum wages would have on enrolhnent and employment. Mattila (197 8)
found that higher minimum wages were associated with higher enrollments of teenagers.
This was due, at least in part, to large disemployment effects among teens formerly not
enrolled in school. These non-enrolled students subsequently return to school.

Ehrenberg and Marcus (1980 and 1982) ﬁnd that minimum wages have mixed
effects, depending on income and race. Higher-income, white, teenage males and lower-
income, black, teenage males tend to leave school for employment at higher minimum

wages, while lower-income, white males return to school. This would suggest some shift

158

in the quality of labor employers seek to hire following an increase in the minimum
wage.

Neumark and Wascher (1996) ﬁnd that higher minimum wages induce teenagers
to leave school. They also ﬁnd that higher minimum wages result in an increase in the
probability that less-skilled teenagers become non-employed and non-enrolled,
potentially making it more difﬁcult for them to ﬁnd successful employment in the future.
This result should be compared with the estimates above, that the expansions of the EITC
not only result in increased employment, but also increased enrollments, due to a strong
reduction in the probability that a single mother between the ages of 17 and 21 is non-
enrolled and non-employed. Similar evidence is presented for single mothers between
ages 17 and 24. These estimates would seem to present strong evidence of the superiority
of the EITC as a policy for assisting low-income parents, especially when compared with
the ﬁndings of Neumark and Wascher on the minimum wage.

The evidence here is broadly consistent with the ﬁndings from the spending
patterns of EITC recipients. A minority of the respondents reported in Smeeding et al.,
and Romich and Weisner claimed to be planning to spend at least a part of their credit on
education. The results here seem to support that claim, although the estimates for school
enrollment obtained here seem larger than would have been expected from the spending

surveys.

159

6.10. CONCLUSIONS AND EXTENTIONS

This study investigates whether the earned income tax credit, while encouraging
labor-market participation, has resulted in decreases in school enrollment. This would be
an indication that the EITC encourages labor-market experience at the cost of lower
general training in school. It is not clear beforehand if one type of human-capital
accumulation is generally better for the long-term economic success of single mothers,
but creating a structure of incentives that subsidizes work, but not school, could lead to a
shift in the composition of the accumulated human capital within this component of the
labor force. The results here strongly suggest that, while the EITC has been associated
with increased labor-force participation, the EITC has also been correlated with increased
school enrollment.

The study does not control for any potential self-selection bias. It is probable that
there are systematic differences between young women who have children before age 22,
and those who do not. These may include home environment, relative value of
education, aptitude for educational success, and economic opportunity. However, the
essential assumption for this study is not that these differences do not exist, but rather that
they have not been changing over time, at least not changing signiﬁcantly between 1987
and 1996.

Structural economic changes have increased the relative importance of education
in labor-market success. Because the responsibility of caring for a child makes it more
difﬁcult for a mother to continue her own education, it is expected that childbearing
would be delayed. Given that women with a higher aptitude for labor-market success

would be disproportionately affected by these economic changes, it would seem

160

reasonable that the remaining women who have children early in their adult years would
be more difﬁcult to move into employment. These women are less skilled, will have
lower educational attainment, and may have less family support to provide child care and
ﬁnancial support. This would tend to bias the results against the EITC, because the pool
of eligible recipients may be becoming less skilled and employable over time.

One possible counter-argument may be that society is generally more accepting
and accommodating to single mothers than it has been in the past. It may also be true
that the societal expectation of work versus public support for single mothers has
changed over time. If these are true, some of the estimated effects of the EITC presented
above may be capturing a shift in behavior based upon a shift in overall societal
expectations.

As the ﬁrst evidence on human capital accumulation and the EITC, this study
opens up additional questions. The snapshot effects on enrollment and work activity
obtained here may or may not allow inference into the long-term effects of the EITC on
economic success. Initially maintaining enrollment status while working may not reﬂect
accurately the amount of learning or skill enhancement being received. In addition,
simply being employed at some point during the year does not imply long-term success in
the labor market. A longitudinal data approach may be best suited to address some of
these questions. For example, a panel data approach that follows a group of women over
many years would allow for comparisons between the level of economic success attained
by women claiming the EITC, and the level attained by childless women.

Another possible avenue to investigate may be to compare the economic success

of women who were eligible for the EITC with those who were not eligible (presumably

161

because they do not have children), and how that relationship has changed over time with
the expansions of the EITC. Examples of the potential areas for investigation would be
ﬁnal educational attainment at some speciﬁc age, asset and wealth accumulation,
employment stability, and earnings.

Finally, additional study into the behavioral and spending patterns of young,
single mothers could provide insight into how programs such as the EITC affect decision
making. Do such programs that encourage work while providing signiﬁcant cash
beneﬁts provide better long-term solutions than simple public support? Instead of
requiring work, the EITC could be structured to provide beneﬁts to single mothers who
succeed in school. Would such a program provide a better long—tenn solution than

requiring work in order to receive beneﬁts?

162

1987
1988
1989
1990
1991
1992
1993
1994
1995
1996

'<
a
an
n

 

1987
1988
1989
1990
1991
1992
1993
1994
1995
1996

Observations

381
341
372
381
368
395
375
370
340
392

Single Women without Children Ages 17-21, 1987-1996

Observations

4,534
4,074
4,382
4,341
4,003
3,834
3,799
3,824
3,433
3,594

Table 6.1
Employment and Enrollment for Single Women Ages 17 - 21

School/Employment Status
Single Women with Children Ages 17-21, 1987-1996

Non-School Non-School
N on-Employed Employed
0.409 0.370
0.378 0.396
0.382 0.390
0.360 0.412
0.389 0.386
0.403 0.339
0.349 0.365
0.289 0.408
0.271 0.459
0.255 0.462
School/Employment Status

Non-School
Non-Employed

0.039
0.047
0.047
0.051
0.048
0.057
0.053
0.060
0.061
0.054

N on-School

163

Employed

0.265
0.251
0.252
0.238
0.221
0.211
0.194
0.203
0.197
0.213

School
Employed

0.1 13
0.109
0.086
0.105
0.095
0.122
0.109
0.149
0.138
0.145

School
Employed

0.508
0.510
0.503
0.502
0.494
0.494
0.510
0.490
0.503
0.491

School
Non-Employed

0.108
0.117
0.142
0.123
0.130
0.137
0.176
0.154
0.132
0.138

School
N on-Employed

0.188
0.192
0.199
0.209
0.237
0.237
0.243
0.246
0.239
0.241

 

1987
1988
1989
1990
1991
1992
1993
1994
1995
1996

Year

1987
1988
1989
1990
1991
1992
1993
1994
1995
1996

Observations

731
685
737
760
734
758
762
747
698
750

Single Women without Children Ages 17-24, 1987-1996

Observations

6,170
5,540
6,029
5,928
5,637
5,426
5,312
5,376
4,650
4,772

Table 6.2
Employment and Enrollment for Single Women Ages 17 - 24

School/Employment Status
Single Women with Children Ages 17-24, 1987-1996

N on-School Non-School
Non-Employed Employed
0.378 0.476
0.366 0.483
0.368 0.479
0.362 0.466
0.369 0.460
0.375 0.429
0.350 0.424
0.301 0.467
0.297 0.500
0.241 0.523
School/Employment Status

Non-School
Non-Employed

0.042
0.045
0.049
0.049
0.053
0.057
0.054
0.060
0.060
0.056

Non-School

164

Employed

0.381
0.360
0.359
0.347
0.334
0.323
0.305
0.313
0.297
0.302

School
Employed

0.078
0.074
0.064
0.086
0.080
0.099
0.113
0.126
0.113
0.140

School
Employed

0.429
0.444
0.437
0.436
0.427
0.434
0.446
0.433
0.448
0.443

School
Non-Employed

0.068
0.076
0.090
0.087
0.090
0.098
0.1 13
0.106
0.090
0.096

School
Non-Employed

0.148
0.152
0.156
0.168
0.186
0.185
0.196
0.194
0.195
0.199

Author

Mattila

Mattila

Ehrenberg and
Marcus

Cunningham

Leighton and
Mincer

Hashimoto

Ehrenberg and
Marcus

Neumark and
Wascher

Neumark and
Wascher

Table 6.3
Summary of Minimum Wage Literature

 

 

Year Data Findings

1979 October CPS Displaced teens return to school after increase
increase in minimum wage

1981 October CPS Full-time employment of teens decreases and
enrolhnent increases after minimum wage
increase.

1980 1966 NLS White teens in lower-income families move

from part-time to full-time, higher-income teens
return to full-time school. Nonwhites move
from school to full-time work.

1970 Census Shift in employment from teens in poor farrriles
to teens in non-poor families.

1981 1960 and 1970 Variation between effects on blacks and whites.
Censuses Probability blacks are employed lower while
higher probability black teenagers are enrolled.

1981 PSID and NLSY Minimum wages lower productivity and wage
growth for men with less schooling.

1982 NLS 1966 and Minimum-wage increase of 1967 resulted in
1969 slower wage growth.

1982 1966 NLS Young With higher minimum wages low-income white
Male and 1969 teens shift from school to work, higher-income
Young Female white teens increase schooling. Black teens
shift from school to work.

1996 Matched Higher minimum wages result in low-skill teens
CPS becoming unemployed, while higher-skill teens
leave school for work at the higher minimum.

2001 1983 and 1991 Minimum wages lower on-the-job training, with the
January CPS reduction acute for formal training. No evidence of
higher pre-employment skill acquisition.

165

Author

Eissa and Hoynes

Smeeding et al.

Romich and
Weisner

Barrow and
McGranahan

2000

2000

2000

Table 6.4
Summary of EITC Literature

Data Findings
March CPS Expansions of the EITC are associated

with reduced labor supply for wives.

Center for Law and Tuition payments listed by 16 percent
Human Services of respondents as one of the top 3 uses
survey of tax ﬁlers of EITC refund.

Survey of New Find evidence that recipients use the
Hope Project EITC for "big-ticket" items, including
participants tuition.

Consumer Evidence that households receiving the

Expenditure EITC save at least some of the reﬁlnd.

Survey

166

Table 6.5
Sample Summary Statistics

Unmarried Unmarried
Women Women
17-24 17-21
Age 20.072 18.843
(2.2608) (1.4088)
Years of Education 14.653 14.429
(3.1893) (3.2927)
Nonwhite 0.196 0.190
(0.3972) (0.3921)
Kids 0.1 18 0.085
(0.3230) (0.2794)
Kids under 6 0.151 0.105
(0.4741) (0.3794)
Enrollment 0.564 0.682
(0.4959) (0.4656)
Employment 0.747 0.709
(0.4345) (0.4542)
NSNE 0.086 0.077
(0.2807) (0.2662)
NSE 0.350 0.241
(0.4770) (0.4277)
SE 0.397 0.468
(0.4893) (0.4990)
SNE 0.166 0.214
(0.3725) (0.4102)
Number of Observations 62,202 43,533

Standard errors in parentheses.

167

Age

Kids

Kids < 6

Uneamed Income

Years of Education

Nonwhite

Nonwhite x Kids

EITC Subsidy Rate

Trend

Number of Observations

Log Likelihood

Table 6.6
Probits on Employment

Single Women
Ages 17-21
0.0759
(0.0019)

.0.2223
(0.0350)

-0.1594
(0.0163)

-0.0059
(0.0009)

0.0117
(0.0009)

-0.2206
(0.0078)

0.1102
(0.0136)

0.7478
(0.1036)

-l.7E-05
(8.39E-5)

43,531
-22,922

Single Women
Ages 17-24
0.0498
(0.0010)

.0.2455
(0.0232)

-0.1295
(0.0084)

-0.0085
(0.0007)

0.0123
(0.0007)

-0.1961
(0.0064)

0.0961
(0.0087)

0.6738
(0.0667)

9.2E—05
(6.57E-5)

62,200
-29,985

Sample of unmarried women from 1987-96. Includes full set of state dummies.
Partial derivatives evaluated at sample means. Standard errors in parentheses.

168

Table 6.7
Probits on Employment
for Single Women Ages 17 through 21

Refundable

State EITCs

0.0746
(0.0019)

-0.2379
(0.0373)

-0.1578
(0.0164)

0.0059
(0.0009)
0.0152
(0.001 1)

-02203
(0.0078)

0.1080
(0.0137)

0.4261
(0.1410)

-0.0338
(0.1354)

-1 .0881
(0.1897)

-1.2088
(0.2524)

0.0344
(0.0290)

0.8529
(0.1303)

With State
EITCs

Age 0.0746
(0.0019)
Kids -0.2444
(0.0367)
Kids < 6 -0.1578
(0.0162)
Unearned Income -0.0059
(0.0009)
Years of Education 0.0152
(0.0011)
Nonwhite -0.2207
(0.0079)
Nonwhite x Kids 0.1110
(0.0136)
Minimum Wage 0.4266
(0.1409)
AFDC -0.0371
(0.1354)
Income Growth -1.0877
(0.1896)
Unemployment Rate -1.2053
(0.2523)
Waivers x Kids -0.0408
(0.0290)
EITC Subsidy Rate 0.9057
(0.1289)
State EITC x Kids -l.8184
(0.4940)

Refundable State EITC

x Kids

Trend -5.5E-05
(9.15E-5)
Number of Observations 43,531
Log Likelihood -22,878

-1 .0989
(0.8926)

-5.4E-05
(9.15E-5)

43,531
-22,889

Sample of unmarried women from 1987-96. Regressions include state dummies.
Partial derivatives evaluated at sample means. Standard errors in parentheses.

169

Table 6.8
Probits on Employment
for Single Women Ages 17 through 24

With State Refundable
EITCs State EITCs
Age 0.0486 0.0487
(0.0010) (0.0010)
Kids -0.2572 -0.2561
(0.0247) (0.0250)
Kids < 6 -0.1281 -0.1278
(0.0084) (0.0084)
Unearned Income -0.0086 -0.0086
(0.0007) (0.0007)
Years of Education 0.0162 0.0162
(0.0009) (0.0009)
Nonwhite -0. 1959 -0. 1956
(0.0064) (0.0064)
Nonwhite x Kids 0.0960 0.0943
(0.0087) (0.0088)
Minimum Wage 0.4734 0.4712
(0.1104) (0.1104)
AFDC -0.1895 -0.1888
(0.1055) (0.1055)
Income Growth -l.0954 -1.0975
(0.1463) (0.1464)
Unemployment Rate -0.8586 -0.8623
(0.1968) (0.1969)
Waivers x Kids -0.0144 -0.0113
(0.0186) (0.0188)
EITC Subsidy Rate 0.7598 0.7389
(0.0831) (0.0845)
State EITC x Kids -1.2562
(0.3576)
Refundable State EITC -0.8989
x Kids (0.6039)
Trend 9.7E-05 9.9E-05
(7.16E-5) (0.0007)
Number of Observations 62,200 62,200
Log Likelihood -29,926 -29,936

Sample of unmarried women from 1987-96. Regressions include state dummies.
Partial derivatives evaluated at sample means. Standard errors in parentheses.

Table 6.9

Multinomial Logits on Employment and Enrollment

for Single Women Ages 17 - 21

No School
Employed
Age 0.3432
(0.0159)
Kids -1 . 1712
(0.1458)
Kids < 6 -0.6018
(0.0768)
Unearned Income -0.2004
(0.0109)
Nonwhite -0.9244
(0.0583)
Nonwhite x Kids 0.5891
(0.1013)
EITC Subsidy Rate 3.2409
(0.5192)
Trend -0.0032
(0.0007)
Constant -4.5287
(0.3094)
Number of Observations 43,533
Log Likelihood -46,781

School
Employed

-0.1782
(0.0152)

-2.9128
(0.2195)

-1.2087
(0.1427)

0.0556
(0.0084)

-1.0767
(0.0538)

1.0323
(0.1279)

3.8938
(0.7087)

-0.0018
(0.0006)

5.9572
(0.2940)

School
99mm

-0.5600
(0.0168)

.2.3570
(0.2011)

-0.5009
(0.1111)

0.0557
(0.0087)

0.0258
(0.0553)

0.4646
(0.1240)

2.6524
(0.7003)

0.0004
(0.0007)

11.7388
(0.3203)

Relative to base case of non-enrolled and non-employed. Sample of unmarried women from 1987-96.
Standard errors in parentheses.

171

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

EITC Subsidy Rate

Trend

Probability

Table 6.10
Partial Derivatives from Table 6.9

No School
Unemployed

-00142
(0.0041)

0.2096
(0.0126)

0.1338
(0.0133)

0.0192
(0.0021)

0.1520
(0.0138)

-0.1074
(0.0141)

-0.6162
(0.0699)

0.0004
(0.0001)

0.2514

No School
Employed

0.1320
(0.0037)

0.2132
(0.0334)

-0.0320
(0.0209)

.0.0572
(0.0023)

-0.1631
(0.0109)

0.0210
(0.0226)

0.3638
(0.1181)

.0.0007
(0.0001)

0.4605

School

Employed

-0.0368
(0.0033)

.0.3232
(0.0571)

-0.1059
(0.0209)

0.0206
(0.0017)

.0.0732
(0.0066)

0.0966
(0.0198)

0.2259
(0.0970)

-9.20E-06
(5 .00E-5)

0.1566

School
Unemployed

-0.081 1
(0.0067)

-0.0996
(0.0408)

0.0041
(0.0126)

0.0174
(0.0016)

0.0842
(0.0090)

-0.0101
(0.0121)

0.0265
(0.0748)

0.0003

(50013-5)

0.1315

Partial derivatives are evaluated at sample means. Assumes woman has one child and is

eligible for an EITC wage subsidy of 34 percent. Standard errors are in parentheses.

172

Multinomial Logits on Employment and Enrollment

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

Minimum Wage

AFDC

Income Growth

Unemployment Rate

Waivers x Kids

EITC Subsidy Rate

State EITC Rate

Trend

Constant

Number of Observations

Log Likelihood

Table 6.11

for Single Women Ages 17 - 21

No School

Employed
0.3491
(0.0160)

-l.0484
(0.1550)

-0.6061
(0.0773)

0.1971
(0.0110)

0.9086
(0.0586)

0.5790
(0.1021)

0.8177
(0.4836)

0.4437
(0.1445)

.3.0077
(1.2113)

-l9.6387
(1.4738)

0.0280
(0.1483)

2.7507
(0.6526)

4.2075
(2.5086)

0.0033
(0.0007)

-2.8170
(0.3979)

43,533
-46,443

School
Employed

0.1726
(0.0153)

2.7523
(0.2296)

-1.2107
(0.1435)

0.0576
(0.0085)

-1.0561
(0.0541)

1.0210
(0.1287)

-1.2971
(0.4680)

0.4224
(0.1384)

4.9114
(1.1693)

-20.7848
(1.4191)

0.0059
(0.1957)

3.2661
(0.8754)

-7.5209
(4.0197)

-0.0020
(0.0007)

8.0106
(0.3817)

School
Unemployed

0.5592
(0.0168)

.2.2577
(0.2118)

0.4936
(0.1110)

0.0574
(0.0088)

0.0183
(0.0555)

0.4722
(0.1247)

0.9173
(0.4980)

0.1766
(0.1465)

—2.0368
(1.2488)

0.1143
(1.5036)

0.2886
(0.1885)

1.8525
(0.8735)

0.4572
(3.1606)

0.0003
(0.0007)

11.7149
(0.4110)

Relative to base case of non-enrolled and non-employed. Sample of unmarried women from 1987-96.

Standard errors in parentheses.

173

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

Minimum Wage

AFDC

Income Growth

Unemployment Rate

Waivers x Kids

EITC Subsidy Rate

State EITC x Kids

Trend

Probability

Table 6.12
Partial Derivatives from Table 6.11

N 0 School
Unemployed

0.0183
(0.0049)

0.2108
(0.0156)

0.1383
(0.0144)

0.0198
(0.0024)

0.1575
(0.0149)

0.1101
(0.0155)

0.1245
(0.0878)

0.0766
(0.0266)

0.6290
(0.2264)

3.3031
(0.3314)

-0.0052
(0.0258)

0.5247
(0.0913)

0.8335
(0.4269)

0.0005
(0.0001 )

0.2614

N 0 School
Employed

0.1298
(0.0044)

0.2080
(0.0367)

-0.0357
(0.0217)

0.0563
(0.0024)

0.1598
(0.0121)

0.0210
(0.0232)

-0.1588
(0.0823)

0.0700
(0.0248)

0.2796
(0.2041)

-3.2560
(0.2958)

-0.0222
(0.0324)

0.3456
0.1461

0.4738
(0.5960)

0.0007
(0.0001)

0.4650

School

Employed

0.0380
(0.0043)

0.3192
(0.0627)

-0.1068
(0.0223)

0.0209
(0.0022)

0.0712
(0.0083)

0.0963
(0.0204)

0.1286
(0.0374)

-0.0203
(0.0107)

0.3926
(0.0910)

-1.2771
(0.1750)

0.0040
(0.0234)

0.1973
0.1196

-0.6790
(0.5186)

-2.58E-05
(5.00E-5)

0.1567

School
Registered

0.0735
(0.0083)

0.0996
(0.0435)

0.0042
(0.01 15)

0.0156
(0.0019)

0.0734
(0.0099)

0.0071
(0.0112)

0.1629
(0.0380)

0.0136
(0.0100)

0.0432
(0.0828)

1.2300
(0.1783)

0.0314
(0.0157)

-0.0181
0.0808

0.3193
(0.3041)

0.0002
(5.00155)

0.1169

Partial derivatives are evaluated at sample means. Assumes woman has one child and is

eligible for an EITC wage subsidy of 34 percent. Standard errors are in parentheses.

174

Age

Kids

Kids < 6

Unearned Income
Nonwhite

Nonwhite x Kids
Minimum Wage
AFDC

Income Growth
Unemployment Rate
Waivers x Kids
EITC Subsidy Rate
Refundable state EITC Rate
Trend

Constant

Number of Observations
Log Likelihood

Table 6.13
Multinomial Logits on Employment and Enrollment
for Single Women Ages 17 - 21

No School
Employed

0.3491
(0.0160)

0.0192
(0.1559)

0.6096
(0.0774)

0.1975
(0.01 10)

-0.9090
(0.0586)

0.5767
(0.1021)

0.8036
(0.4836)

0.4608
(0.1449)

-2.9804
(1.2114)

-19.4691
(1.4711)

0.0016
(0.1481)

2.5304
(0.6547)

2.0897
(4.5487)

0.0032
(0.0007)

-2.8287
(0.3979)

43,533
-46,443

School
Employed

0.1727
(0.0153)

-2.7630
(0.2304)

0.2073
(0.1437)

0.0577
(0.0085)

-l.0564
(0.0541)

1.0048
(0.1287)

-1.2859
(0.4679)

0.4297
(0.1386)

-4.8736
(1.1694)

-20.6378
(1.4171)

-0.0091
(0.1966)

3.2934
(0.8822)

-16.1461
(9.5733)

0.0019
(0.0007)

7.9984
(0.3816)

School
Unemployed

-0.5592
(0.0168)

.2.2579
(0.2130)

0.4934
(0.1 1 14)

0.0574
(0.0088)

0.0181
(0.0555)

0.4715
(0.1248)

0.9246
(0.4980)

-0.1838
(0.1468)

.2.0112
(1.2490)

0.9958
(1.5014)

0.2867
(0.1886)

1.8512
(0.8765)

0.0569
(6.5143)

0.0003
(0.0007)

11.7056
(0.4109)

Relative to base case of non-enrolled and non-employed. Sample of unmarried women from 1987-96.

Standard errors in parentheses.

175

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

Minimum Wage

AFDC

Income Growth

Unemployment Rate

Waivers x Kids

EITC Subsidy Rate

Refundable State EITC

x Kids

Trend

Probability

Table 6.14

No School
Unemployed

0.0170
(0.0050)

0.2159
(0.0158)

0.1413
(0.0147)

0.0196
(0.0025)

0.1583
(0.0150)

-0.1 l 16
(0.0157)

0.1229
(0.0889)

0.0801
(0.0270)

0.6338
(0.2293)

3.3048
(0.3333)

0.0086
(0.0262)

0.5065
(0.0946)

0.4729
(0.8563)

0.0005
(0.0001)

0.2672

No School
Employed

0.1293
(0.0044)

0.2119
(0.0353)

0.0366
(0.0217)

0.0562
(0.0024)

0.1583
(0.0121)

0.0222
(0.0231)

0.1557
(0.0820)

0.0730
(0.0247)

-0.2756
(0.2033)

-3.2165
(0.2953)

0.0152
(0.0323)

0.2876
(0.1467)

1.7483
(1.1749)

0.0007
(0.0001)

0.4530

Partial Derivatives from Table 6.13

School

Employed

0.0379
(0.0044)

0.3279
(0.0632)

-0. 1089
(0.0227)

0.0210
(0.0023)

-0.0737
(0.0086)

0.0960
(0.0205)

0.1325
(0.0382)

0.0209
(0.0109)

-0.4013
(0.0928)

0.3266
(0.1792)

0.0066
(0.0241)

0.2242
(0.1247)

-2.3063
(1.3016)

-3.06E-05
(50013-5)

0.1604

School
9mm

0.0744
(0.0084)

0.0999
(0.0446)

0.0042
(0.0118)

0.0156
(0.0019)

0.0737
(0.0099)

0.0066
(0.01 14)

0.1653
(0.0387)

0.0138
(0.0102)

0.0431
(0.0844)

1.2384
(0.1805)

0.0304
(0.0161)

-0.0053
(0.0837)

0.0851
(0.6623)

0.0002

(50015-5)

0.1194

Partial derivatives are evaluated at sample means. Assumes woman has one child and is

eligible for an EITC wage subsidy of 34 percent. Standard errors are in parentheses.

176

Table 6.15
Estimated Changes in Probabilities for Women
with Two or More Children

Case 1 -- Increase in subsidy rate from 40 to 45 percent for women with
2 or more children

 

 

NSNE NSE SE SNE
Subsidy at 40 percent 0.317 0.242 0.103 0.338
Subsidy at 45 percent 0.292 0.255 0.111 0.342
Change in probability -0.025 0.014 0.009 0.003

Case 2 -- Change in probabilities for women with and without children from
1992 to 1996

w E§E EL? $2

Women with 2 children

1992 0.468 0.205 0.076 0.251

1996 0.344 0.269 0.112 0.275
Difference -0.124 0.064 0.036 0.024
Women without children

1992 0.034 0.087 0.542 0.337

1996 0.035 0.087 0.540 0.339
Difference 0.000 -0.001 -0.002 0.002

 

 

 

 

 

 

177

Table 6.16
Estimated Changes in Probabilities for Women with One Child

Case 1 -- Increase in subsidy rate from 34 to 40 percent for women with

 

 

 

1 child
N SNE NSE SE SNE
Subsidy at 34 percent 0.215 0.255 0.193 0.337
Subsidy at 40 percent 0.191 0.267 0.208 0.334
Change in probability -0.024 0.012 0.015 0003

Case 2 -- Change in probabilities for women with and without children from
1992 to 1996

m _N_$_I‘.3. S_E .S_.1‘LE_

Women with 1 child

1992 0.314 0.247 0.167 0.272

1996 0.233 0.284 0.210 0.273
Difference -0.081 0.037 0.043 0.002
Women without children

1992 0.047 0.118 0.731 0.105

1996 0.047 0.1 17 0.729 0.107
Difference 0.000 -0.001 -0.001 0.002

 

 

 

 

 

 

178

Table 6.17

Multinomial Logits on Employment and Enrollment

for Single Women Ages 17 - 24

No School
Employed
Age 0.2638
(0.0080)
Kids -1.2124
(0.0976)
Kids < 6 -0.6197
(0.043 1)
Unearned Income -0.1708
(0.0066)
Nonwhite -0.9465
(0.0473)
Nonwhite x Kids 0.6505
(0.0743)
EITC Subsidy Rate 2.3402
(0.3567)
Trend -0.0027
(0.0005)
Constant -3.0280
(0.1685)
Number of Observations 62,202
Log Likelihood -64,499

School
Employed

0.1834
(0.0081)

-2.8453
(0.1559)

0.1355
(0.0895)

0.031 1
(0.0053)

0.0203
(0.0460)

0.8612
(0.1006)

4.0714
(0.5252)

0.0006
(0.0005)

5.9799
(0.1692)

School
Unemployed

0.4552
(0.0097)

-2. 1926
(0.1609)

0.4679
(0.081 1)

0.0432
(0.0057)

0.0722
(0.0482)

0.2221
(0.1038)

2.4192
(0.5662)

0.0014
(0.0006)

9.7951
(0.1959)

Relative to base case of non-enrolled and non-employed. Sample of unmarried women from 1987-96.
Standard errors in parentheses.

179

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

EITC Subsidy Rate

Trend

Probability

Table 6.18
Partial Derivatives from Table 6.17

N 0 School
Unemployed

0.0139
(0.0022)

0.2119
(0.0092)

0.1364
(0.0082)

0.0190
(0.0013)

0.1644
(0.0114)

0.1071
(0.0110)

0.5228
(0.0531)

0.0003
(0.0001)

0.2589

No School
Employed

0.1016
(0.0020)

0.1887
(0.0273)

0.0447
(0.0125)

-0.0471
(0.0014)

-0.l709
(0.0088)

0.0680
(0.0165)

0.1549
(0.0841)

-0.0007
(8.00E-5)

0.4832

School

Employed

0.0380
(0.0022)

0.3397
(0.0402)

-0.0975
(0.0133)

0.0167
(0.0010)

0.0657
(0.0048)

0.0644
(0.0144)

0.3288
(0.0777)

0.0001

(4.00135)

0.1602

School
Unemployed

0.0497
(0.0036)

0.0610
(0.0239)

0.0058
(0.0071)

0.01 14
(0.0009)

0.0721
(0.0067)

-0.0253
(0.0073)

0.0390
(0.0489)

0.0003

(4.00135)

0.0977

Partial derivatives are evaluated at sarrrple means. Assumes woman has one child and is

eligible for an EITC wage subsidy of 34 percent. Standard errors are in parentheses.

180

Multinomial Logits on Employment and Enrollment

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

Minimum Wage

AFDC

Income Growth

Unemployment Rate

Waivers x Kids

EITC Subsidy Rate

State EITC Rate

Trend

Constant

Number of Observations

Log Likelihood

Table 6.19

Single Women Ages 17 - 24

No School School
Employed Employed
0.2693 -0.1784
(00080) (0.0081)
-1.0926 -2.6038
(0.1041) (0.1642)
-0.6210 -1.1301
(0.0434) (0.0899)
-0.1675 0.0335
(0.0067) (0.0054)
-0.9330 -1 .0047
(0.0475) (0.0462)
0.6240 0.8434
(0.0749) (0.1011)
-0.6141 -0.9223
(0.3828) (0.3886)
-0.5498 -0.5670
(0.1137) (0.1146)
-2.6848 -4.8632
(0.9600) (0.9681)
-17.9694 -l8.2522
(1.1611) (1.1751)
0.0356 0.2990
(0.1022) (0.1430)
1.7622 2.7254
(0.4470) (0.6578)
-1.6188 -6.9164
(1.9221) (3.3712)
-0.0026 -0.0008
(0.0005) (0.0005)
-1.4707 7.8084
(0.2607) (0.2639)
62,202
-64,1 10

School
Unemployed

0.4545
(0.0097)

.2.1391
(0.1694)

0.4634
(0.081 1)

0.0450
(0.0058)

0.0639
(0.0483)

0.2178
(0.1043)

1.0202
(0.4260)

0.1865
(0.1248)

-2.2306
(1.0648)

0.0754
(1.2839)

0.1631
(0.1547)

1.9684
(0.7004)

0.0836
(2.7592)

0.0012
(0.0006)

9.6872
(0.2951)

Relative to base case of non-enrolled and non-employed. Sample of unmarried women from 1987-96.

Standard errors in parentheses.

181

Table 6.20
Partial Derivatives from Table 6.19

No School N 0 School School School
Unemployed Employed Employed Unemployed
Age -0.0172 0.1007 -0.0341 -0.0493
(0.0028) (0.0025) (0.0028) (0.0047)
Kids 0.2144 0.1606 -0.2885 -0.0865
(0.01 1 1) (0.0307) (0.0448) (0.0296)
Kids < 6 0.1400 -0.0557 -0.0884 0.0041
(0.0089) (0.0128) (0.0132) (0.0070)
Unearned Income 0.0200 -0.0462 0.0150 0.0112
(0.0015) (0.0014) (0.0013) (0.0011)
Nonwhite 0.1675 -0.1751 -0.0588 0.0663
(0.0121) (0.0095) (0.0056) (0.0076)
Nonwhite x Kids -0. 1087 0.0715 0.0598 -0.0227
(0.0121) (0.0166) (0.0135) (0.0075)
Minimum Wage 0.0918 -0.1371 -0.0837 0.1290
(0.0732) (0.0665) (0.0287) (0.0282)
AFDC 0.1011 -0.0896 -0.0286 0.0171
(0.0222) (0.0201) (0.0086) (0.0077)
Income Growth 0.6106 -0.2319 -0.3765 -0.0022
(0.1878) (0.1662) (0.0715) (0.0624)
Unemployment Rate 3.1605 -3.1806 -0.9674 0.9876
(0.2574) (0.2267) (0.1175) (0.1258)
Waivers x Kids -0.0208 -0.0193 0.0317 0.0084
(0.0192) (0.0227) (0.0142) (0.0120)
EITC Subsidy Rate 03958 0.1610 0.1835 0.0513
(0.0715) 0.1025 0.0844 0.0606
State EITC x Kids 0.4867 0.0686 -0.7311 0.1759
(0.3536) (0.4596) (0.4052) (0.2228)
Trend 0.0004 -0.0007 6.95E-05 0.0002
(0.0001) (9.00E-5) (4.00E-5) (4.00E-5)
Probability 0.2765 0.4863 0.1418 0.0954

Partial derivatives are evaluated at sample means. Assumes woman has one child and is
eligible for an EITC wage subsidy of 34 percent. Standard errors are in parentheses.

182

Multinomial Logits on Employment and Enrollment

Age

Kids

Kids < 6

Unearned Income
Nonwhite

Nonwhite x Kids
Minimum Wage
AFDC

Income Growth
Unemployment Rate
Waivers x Kids
EITC Subsidy Rate
Refundable state EITC Rate
Trend

Constant

Number of Observations
Log Likelihood

Table 6.21

Single Women Ages 17 - 24

No School School
Employed Employed
0.2695 -0.1783
(0.0080) (0.0081)
-1.0812 -2.6083
(0.1050) (0.1652)
-0.6218 -1.1297
(0.0434) (0.0901)
01676 0.0335
(0.0067) (0.0054)
-0.9332 -1.0049
(0.0475) (0.0462)
0.6231 0.8340
(0.0749) (0.1012)
-0.6086 -0.9163
(0.3828) (0.3885)
-0.5620 -0.5759
(0.1 140) (0.1 149)
-2.6776 -4.8447
(0.9604) (0.9684)
-17.8813 -18.1546
(1.1590) (1.1734)
0.0489 0.3046
(0.1024) (0. 143 7)
1.6686 2.7062
(0.4528) (0.6663)
1.0981 -8.6026
(3.1719) (6.2120)
-0.0026 -0.0008
(0.0005) (0.0005)
-1.4766 7.7990
(0.2606) (0.2638)
62,202
-64,1 1 1

School
Unemployed

0.4544
(0.0097)

-2.1256
(0.1707)

0.4674
(0.0813)

0.0449
(0.0058)

0.0637
(0.0483)

0.2201
(0.1043)

1.0239
(0.4260)

-0.1967
(0.1251)

-2.2268
(1.0650)

0.9982
(1.2821)

0.1721
(0.1547)

1.8941
(0.7047)

2.9507
(4.6815)

0.0012
(0.0006)

9.6820
(0.2951)

Relative to base case of non-enrolled and non-employed. Sample of unmarried women from 1987-96.

Standard errors in parentheses.

183

Age

Kids

Kids < 6

Unearned Income

Nonwhite

Nonwhite x Kids

Minimum Wage

AFDC

Income Growth

Unemployment Rate

Waivers x Kids

EITC Subsidy Rate

Refundable State EITC

x Kids

Trend

Probability

Table 6.22

No School
Unemployed

0.0172
(0.0029)

0.2169
(0.0112)

0.1415
(0.0090)

0.0201
(0.0016)

0.1685
(0.0122)

0.1094
(0.0123)

0.0915
(0.0737)

0.1041
(0.0225)

0.6142
(0.1894)

3.1663
(0.2589)

0.0233
(0.0194)

-0.3838
(0.0736)

0.1155
(0.6029)

0.0004
(0.0001)

0.2801

No School
Employed

0.1004
(0.0025)

0.1614
(0.0307)

0.0563
(0.0129)

0.0462
(0.0014)

0.1749
(0.0095)

0.0726
(0.0166)

0.1361
(0.0666)

0.0918
(0.0202)

0.2339
(0.1666)

-3.1735
(0.2271)

0.0166
(0.0228)

0.1440
(0.1037)

0.7289
(0.7834)

-0.0007

(9.00E-5)

0.4826

Partial Derivatives from Table 6.21

School

Employed

0.0341
(0.0028)

0.2935
(0.0453)

-0.0888
(0.0133)

0.0150
(0.0013)

0.0593
(0.0057)

0.0589
(0.0135)

~0.0838
(0.0288)

0.0290
(0.0086)

0.3770
(0.0718)

0.9737
(0.1185)

0.0315
(0.0143)

0.1899
(0.0862)

0.1644
(0.7479)

6.90E-05
(40013-5)

0.1422

School
Unemployed

-0.0491
(0.0047)

-0.0848
(0.0296)

0.0036
(0.0070)

0.0111
(0.0011)

0.0657
(0.0075)

0.0221
(0.0075)

0.1285
(0.0282)

0.0167
(0.0077)

0.0032
(0.0623)

0.9808
(0.1257)

0.0085
(0.01 19)

0.0499
(0.0608)

0.3200
(0.3844)

0.0002

(4.00E-5)

0.0952

Partial derivatives are evaluated at sarrrple means. Assumes woman has one child and is

eligible for an EITC wage subsidy of 34 percent. Standard errors are in parentheses.

184

APPENDIX

185

State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii

Idaho

Illinois
Indiana

Iowa

Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada

New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas

Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming

U.S. Total

 

Table A-1
EITC and Federal Tax Statistics by State - 1998

 

 

 

 

EITC EITCs as % Returns EITCs as %
lie—dam Mg won—laden Ei|_9<1 22mm
443,441 4,351,037 10.2% 1,879,778 23.6%
30,833 615,205 5.0% 328,972 9.4%
342,692 4,667,277 7.3% 2,037,046 16.8%
250,174 2,538,202 9.9% 1,095,998 22.8%
2,373,809 32,682,794 7.3% 14,196,991 16.7%
235,744 3,968,967 5.9% 1,973,524 11.9%
146,219 3,272,563 4.5% 1,625,515 9.0%
49,796 744,066 6.7% 364,216 13.7%
52,638 521,426 10.1% 272,169 19.3%
1,283,199 14,908,230 8.6% 7,076,098 18.1%
702,954 7,636,522 9.2% 3,471,292 20.3%
67,211 1,190,472 5.6% 553,525 12. 1%
83,836 1,230,923 6.8% 533,886 15.7%
753,848 12,069,774 6.2% 5,636,139 13.4%
364,171 5,907,617 6.2% 2,765,157 13.2%
148,997 2,861,025 5.2% 1,335,353 11.2%
146,827 2,638,667 5.6% 1,200,088 12.2%
303,911 3,934,310 7.7% 1,694,158 17.9%
475,701 4,362,758 10.9% 1,847,098 25.8%
80,815 1,247,554 6.5% 584,019 13.8%
324,565 5,130,072 6.3% 2,459,510 13.2%
280,285 6,144,407 4.6% 3,011,463 9.3%
562,121 9,820,231 5.7% 4,497,948 12.5%
217,542 4,726,411 4.6% 2,298,213 9.5%
343,825 2,751,335 12.5% 1,153,947 29.8%
385,385 5,437,562 7.1% 2,492,001 15.5%
65,948 879,533 7.5% 409,958 16.1%
94,016 1,660,772 5.7% 795,895 11.8%
125,729 1,743,772 7.2% 873,591 14.4%
56,003 1,185,823 4.7% 598,913 9.4%
446,071 8,095,542 5.5% 3,924,125 11.4%
179,459 1,733,535 10.4% 767,985 23.4%
1,302,422 18,159,175 7.2% 8,268,276 15.8%
642,853 7,545,828 8.5% 3,516,043 18.3%
36,849 637,808 5.8% 300,297 12.3%
687,933 11,237,752 6.1% 5,472,752 12.6%
283,049 3,339,478 8.5% 1,435,811 19.7%
202,177 3,282,055 6.2% 1,522,027 13.3%
692,596 12,002,329 5.8% 5,654,094 12.2%
58,782 987,704 6.0% 472,064 12.5%
373,985 3,839,578 9.7% 1,748,410 21.4%
48,174 730,789 6.6% 344,408 14.0%
479,331 5,432,679 8.8% 2,491,859 19.2%
1,879,486 19,712,389 9.5% 8,693,863 21.6%
108,892 2,100,562 5.2% 896,348 12.1%
36,032 590,579 6.1% 287,782 12.5%
441,337 6,789,225 6.5% 3,197,582 13.8%
311,202 5,687,832 5.5% 2,669,856 11.7%
138,238 1,811,688 7.6% 739,784 18.7%
249,342 5,222,124 4.8% 2,513,562 9.9%
32,441 480,045 6.8% 229,223 14.2%
19,442,159 270,248,003 7.2% 125,393,584 15.5%

Source: Statistics of Income, lntemal Revenue Service.

186

 

Year EITC
State adopted
Maryland 1987
Vermont 1988
Wisconsin 1989
Iowa 1990
Minnesota 1991
New York 1994
Massachusetts 1997
Oregon 1997
Kansas 1998
Colorado 1999
Illinois 2000
Missouri 2000
New Jersey 2000
District of Columbia 2000
Average
Average without NY or DC

Source: Bureau of Labor Statistics, U.S. Department of Labor.

Table A-2
Unemployment Rate

Year before
adoption

-2.5%
-2.6%
-l.2%
-l.0%
-0.7%
0.9%
-1.1%
0.5%
-1. 1%
-0.7%
0.1%
-0.8%
0.4%
2.1%

-0.6%
-0.9%

187

Two years
before adoption

-2.6%
-2.5%
-0.7%
-l 0%
-0.9%

1.0%
-0.7%
-0.2%
-1.0%
-l.2%

0.0%
-0.6%

0.3%

3.2%

-0.5%
-0.9%

Year after
ail—012m.

-l.0%
-l.6%
-l.2%
-2.2°/o
-2.3%

0.7%
-l.2%

1.1%
-l.2%

-l.0%
-l.2%

Two years
after adoption

-l.3%
-l.1%
-l.3%
-2.5%
-2.1%

0.8%
-1.1%

1.3%

-0.9%
-l .1%

m adopted
Maryland 1987
Vermont 1988
Wisconsin 1989
Iowa 1990
Minnesota 1991
New York 1994
Massachusetts 1997
Oregon 1997
Kansas 1998
Colorado 1999
Illinois 2000
Missouri 2000
New Jersey 2000
District of Columbia 2000
Average

Average without NY or DC

Source: Census Bureau, U.S. Department of Commerce.

Table A-3
Poverty Rate

Year EITC Year before

adoption

-4.8%
-4.1%
-5.2%
-2.5%
-1.5%

1.3%
-3.6%
-1.9%
-3.6%
-3.5%
-1.9%
-0.2%
-4.0%

3.1%

-2.3%
3.1%

188

Two years
before adoption

-4.9%
-3.6%
-4.8%
-3. 1%
-1.6%
1.1%
-3.2%
-2.3%
-3.1%
-4.3%
-2.3%
-1.6%
-4.1%
6.4%

-2.2%
-3.2%

Year after
adopﬁon

-3.2%
-4.8%
-4.2%
-4.6%
-1.8%
2.7%
-4.0%
2.3%
0.4%

-1.9%
-2.5%

Two years
MM

-3.5%
-3.7%
-4.3%
-4.0%
-2.7%

2.9%
-2. 1%

1.6%

-2.0%
-2.7%

Table A—4
Labor Force Participation

Year EITC Year before

§t_a_t_e adopted
Maryland 1987
Vermont 1988
Wisconsin 1989
Iowa 1990
Minnesota 1991
New York 1994
Massachusetts 1997
Oregon 1997
Kansas 1998
Colorado 1999
Illinois 2000
Missouri 2000
New Jersey 2000
District of Columbia 2000
Average

Average without NY or DC

adopﬁon

4.0%
5.8%
4.5%
3.7%
6.5%
-4.6%
0.7%
-1.2%
3.4%
7.4%
2.6%
1.6%
0.0%
0.6%

2.5%
3.3%

Two years
before adoption

3.5%
6.1%
3.9%
3.8%
5.9%
4.6%
0.8%
1.9%
3.3%
6.5%
2.0%
2.0%
-0.3%
-1.0%

2.4%
3.3%

Source: Bureau of Labor Statistics, U.S. Department of Labor.

189

Year after
adopﬁon

3.9%
4.9%
3.8%
3.9%
6.2%
-5.7%
1.8%
2.0%
5.0%

2.9%
3.9%

Two years
after adoption

4.2%
4.6%
3.8%
1.8%
6.8%
-5.4%
1.7%
1.6%

2.4%
3.5%

Table A-4
Labor Force Participation

Year EITC Year before

 

State adopted
Maryland 1987
Vermont 1988
Wisconsin 1989
Iowa 1990
Minnesota 199 1
New York 1994
Massachusetts 1997
Oregon 1997
Kansas 1998
Colorado 1999
Illinois 2000
Missouri 2000
New Jersey 2000
District of Columbia 2000
Average

Average without NY or DC

adopﬁon

4.0%
5.8%
4.5%
3.7%
6.5%
—4.6%
0.7%
-1.2%
3.4%
7.4%
2.6%
1.6%
0.0%
0.6%

2.5%
3.3%

Two years

before adoption

3.5%
6.1%
3.9%
3.8%
5.9%
-4.6%
0.8%
1.9%
3.3%
6.5%
2.0%
2.0%
-0.3%
-1.0%

2.4%
3.3%

Source: Bureau of Labor Statistics, U.S. Department of Labor.

189

Year after
adopﬁon

3.9%
4.9%
3.8%
3.9%
6.2%
-5.7%
1.8%
2.0%
5.0%

2.9%
3.9%

Two years
W

4.2%
4.6%
3.8%
1.8%
6.8%
-5.4%
1.7%
1.6%

2.4%
3.5%

Table A—5
AFDC/T AN F Caseloads Per 1,000 People

Year EITC Year before

 

State adopted
Maryland 1987
Vermont 1988
Wisconsin 1989
Iowa 1990
Minnesota 1991
New York 1994
Massachusetts 1997
Oregon 1997
Kansas 1998
Colorado 1999
Illinois 2000
Missouri 2000
New Jersey 2000
District of Columbia 2000
Average

Average without NY or DC

adopﬁon

0.2
-1.2

3.9
-2.3
-2.7

4.6
-2.4
—6.1
-6.8
-6.1

0.9
-0.2
-1.9
27.8

0.5
-2.1

Two years
before ﬂoutiog

0.7
-1.0

4.5
-1.9
-2.6

4.1
-l.9
-6.0
-7.1
-6.6

1.7
-0.4
-2.1
28.7

0.7
-1.9

Year after
adoption

-1.62
-2.37

0.73
-4.30
-4.01

6.78
-0.93
—6.49
-4.91

-1.90
-2.99

Two years
after ado tion

-1.66
-2. 18

0.11
-4.70
-4.43

6.84
-0.84
-5.62

-1.56
-2.76

Sources: U.S. Department of Health and Human Services and Census Bureau, U.S. Department

of Commerce.

190

TABLE A-6

Probits Using Combined Federal and State Subsidy Rate

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Some college

Degree

Low education x Kids
Nonwhite x low education
Non labor income
Subsidy rate

AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Labor-force
Participation

0.0134
-0.0002
-0.0687
-0.0250
-0.0500

0.0140
-0.1534

0.0411

0.0404

0.0042
~0.0112
-0.0029

0.1018
8.7E-05
-1.0629

0.1051
-0.2492

0.0100
-0.1 192

0.0019
-0.0086
-0.0172
-0.0272
-0.0356
-0.0312
-0.0281
-0.0456
-0.0499
-0.0667

83,935
-25,767

(0.0018)

(2.41505)

(0.0066)
(0.0016)
(0.0025)
(0.0048)
(0.0051)
(0.0043)
(0.0034)
(0.0048)
(0.0057)
(0.0001)
(0.0223)

(1.01305)

(0.1347)
(0.081 1)
(0.0953)
(0.0069)
(0.0724)
(0.0061)
(0.0084)
(0.0079)
(0.0105)
(0.0107)
(0.0094)
(0.0073)
(0.0085)
(0.01 1 1)
(0.0124)

Below

Poverty

-0.0085
5.7E-05
0.0581
0.0712
0.0695
0.0159
0.2243
-0.0511
-0.0859
0.01 10
0.0373

-O.1057
-8.2E-07
0.7179
0.0715
0.1855
-0.0193
0.0357
-0.0012
0.0076
0.0143
0.0016
0.0026
0.0030
0.0177
0.0356
0.0262
0.0496

83,935
-32,570

(0.0024)

(3.3E-05)

(0.0078)
(0.0024)
(0.0037)
(0.0075)
(0.0062)
(0.0055)
(0.0042)
(0.0065)
(0.0081)

(0.0295)

(1.4505)

(0.1766)
(0.1074)
(0.1235)
(0.0946)
(0.0993)
(0.0082)
(0.0108)
(0.0099)
(0.0119)
(0.0119)
(0.0105)
(0.0089)
(0.0099)
(0.0123)
(0.0137)

Standard errors in parentheses. All regressions include full set of state dummies.
Estimates are partial derivatives evaluated at the sample means.

191

Below 150%
of Povertv Level

-0.0135
0.0001
0.0868
0.1059
0.0855
0.0200
0.2860

-0.0828

-0.1345
0.0209
0.0421

-0.1012
-2.0E-05
0.8562
0.3292
0.2456
-0.0089
0.2304
-0.0102
0.0012
0.0088
0.0036
0.0056
-0.0035
0.0095
0.0382
0.0419
0.0586

83,935
-39,395

(0.0030)

(4.1E-05)

(0.0091)
(0.0036)
(0.0054)
(0.0094)
(0.0069)
(0.0065)
(0.0051)
(0.0080)
(0.0097)

(0.0374)

(1.8E-05)

(0.2209)
(0.1364)
(0.1533)
(0.0122)
(0.1214)
(0.0103)
(0.0132)
(0.0121)
(0.0150)
(0.0149)
(0.0131)
(0.0107)
(0.0119)
(0.0149)
(0.0160)

TABLE A-7

Probits Using Separate Federal and State Subsidy Rate

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Some college

Degree

LOW education x Kids
Nonwhite x low education
Non labor income
Federal subsidy rate
State subsidy rate
AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. A11 regressions include full set of state dummies.

Labor-force
Participation

0.0133
-0.0002
-0.0686
-0.0251
-0.0500

0.0140
-0.1532

0.0410

0.0405

0.0038
-0.01 l 1
-0.0029

0.1222
~0.1373
8.7E-05
-1.0309

0.0792
-0.2465

0.0066
-0.1210

0.0018
-0.0083
-0.0169
-0.0268
-0.0351
-0.0306
-0.0283
-0.0453
-0.0502
-0.0665

83,935
-25,764

(0.0018)

(2,413-05)

(0.0066)
(0.0016)
(0.0025)
(0.0048)
(0.0051)
(0.0043)
(0.0034)
(0.0048)
(0.0057)
(0.0001)
(0.0244)
(0.1143)

(1.0E-05)

(0.1354)
(0.0820)
(0.0952)
(0.0071)
(0.0724)
(0.0061)
(0.0084)
(0.0079)
(0.0105)
(0.0107)
(0.0094)
(0.0073)
(0.0084)
(0.01 1 1)
(0.0124)

Below

Poverty

-0.0084
5.7E-05
0.0581
0.0713
0.0696
0.0159
0.2243
-0.0510
-0.0859
0.0113
0.0372

-0.1195
0.0664
-1.lE-06
0.6966
0.0910
0.1846
-0.0170
0.0367
-0.0011
0.0074
0.0141
0.0014
0.0023
0.0027
0.0178
0.0354
0.0263
0.0495

83,935
—32,570

Estimates are partial derivatives evaluated at the sample means.

192

(0.0024)

(3.3E-05)

(0.0078)
(0.0024)
(0.0038)
(0.0076)
(0.0062)
(0.0055)
(0.0042)
(0.0065)
(0.0081)

(0.0318)
(0.1972)

(1.4505)

(0.1774)
(0.1083)
(0.1235)
(0.0098)
(0.0993)
(0.0082)
(0.0108)
(0.0099)
(0.01 19)
(0.01 19)
(0.0105)
(0.0089)
(0.0099)
(0.0123)
(0.0137)

Below 150%
of Poverty Level

-0.0135
0.0001
0.0868
0.1060
0.0856
0.0200
0.2860

-0.0828

-0.1345
0.0211
0.0420

-0.1133
0.0521
-2.0E-05
0.8386
0.3470
0.2451
-0.0069
0.2314
-0.0102
0.0010
0.0087
0.0034
0.0053
-0.0038
0.0095
0.0381
0.0420
0.0584

83,935
-39,395

(0.0030)

(4.1E-05)

(0.0091)
(0.0036)
(0.0054)
(0.0094)
(0.0069)
(0.0065)
(0.0051)
(0.0080)
(0.0097)

(0.0403)
(0.2690)

(1.8E-05)

(0.2216)
(0.1371)
(0.1533)
(0.0126)
(0.1214)
(0.0103)
(0.0132)
(0.0121 )
(0.0150)
(0.0149)
(0.0131)
(0.0107)
(0.0119)
(0.0149)
(0.0160)

TABLE A-8

Probits Using Separate Federal and Refundable State Rate

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Some college

Degree

Low education x Kids
Nonwhite x low education
Non labor income
Federal subsidy rate
Refundable state EITC
AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. All regressions include full set of state dummies.

Labor-force
Participation

0.0133
-0.0002
-0.0685
-0.0251
-0.0500

0.0138
-0. 1533

0.0410

0.0405

0.0038
-0.01 1 1
-0.0029

0.1 174
-0.0490

-8.8E-05
-l.0430

0.0881
-0.2469

0.0074
-0.1 193

0.0018
-0.0085
-0.0170
-0.0270
-0.0354
-0.0309
-0.0284
-0.0456
-0.0506
-0.0671

83,935
-25,765

(0.0018)

(2.41305)

(0.0066)
(0.0016)
(0.0025)
(0.0048)
(0.0051)
(0.0043)
(0.0034)
(0.0048)
(0.0057)
(0.0001)
(0.0244)
(0.1710)

(1.01305)

(0.1353)
(0.0819)
(0.0953)
(0.0071)
(0.0725)
(0.0061)
(0.0084)
(0.0079)
(0.0105)
(0.0107)
(0.0094)
(0.0073)
(0.0085)
(0.0112)
(0.0124)

Below

Poverty

-0.0084
5.7E-05
0.0580
0.0712
0.0696
0.0161
0.2243
-0.0510
-0.0859
0.0114
0.0372

-0.1219
0.1770
-1 .2E-06
0.6931
0.0949
0.1806
-0.0162
0.0356
-0.0010
0.0076
0.0142
0.0015
0.0023
0.0027
0.0179
0.0354
0.0263
0.0493

83,935
-32,569

Estimates are partial derivatives evaluated at the sample means.

193

(0.0024)

(3313-05)

(0.0078)
(0.0024)
(0.0038)
(0.0076)
(0.0062)
(0.0055)
(0.0042)
(0.0065)
(0.0081)

(0.0316)
(0.2444)

(1.4505)

(0.1773)
(0.1085)
(0.1235)
(0.0097)
(0.0993)
(0.0082)
(0.0108)
(0.0099)
(0.0119)
(0.01 19)
(0.0105)
(0.0089)
(0.0099)
(0.0122)
(0.0137)

Below 150%
of Povertj Level

-0.0135
0.0001
0.0867
0.1060
0.0856
0.0202
0.2860

-0.0827

-0.1345
0.0212
0.0420

-0.1 149
0.1371
-2.0E-05
0.8363
0.3500
0.2424
-0.0063
0.2307
-0.0101
0.0012
0.0087
0.0035
0.0054
-0.0038
0.0096
0.0380
0.0420
0.0583

83,935
-39,395

(0.0030)

(4.1E-05)

(0.0091)
(0.0036)
(0.0054)
(0.0094)
(0.0069)
(0.0065)
(0.0051)
(0.0080)
(0.0097)

(0.0399)
(0.3318)

(1.8E-05)

(0.2216)
(0.1378)
(0.1533)
(0.0125)
(0.1214)
(0.0103)
(0.0132)
(0.0121)
(0.0150)
(0.0149)
(0.0131)
(0.0107)
(0.0119)
(0.0149)
(0.0160)

Probits Using Combined Federal and State Subsidy Rate with

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Non labor income
Subsidy rate

AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. A11 regressions include ﬁill set of state dummies.

TABLE A-9

Low-Education Subsample

Labor-force
Participation

0.0176
-0.0002
-0.1 102
-0.0365
-0.0754

0.0097
-0.1977
-0.0057

0.2239
1.6E-04
-1.4283

0.2777
-0.3493

0.0028
-0.1682
-0.001 1
-0.0158
-0.0249
-0.0390
-0.0493
-0.0494
-0.0593
—0.0749
-0.0780
-0. 1010

41,708
-17,016

(0.0034)

(4.6E-05)

(0.0100)
(0.0029)
(0.0046)
(0.0098)
(0.0057)
(0.0005)
(0.0440)

(2.0E-05)

(0.2524)
(0.1447)
(0.1857)
(0.0132)
(0.1396)
(0.01 13)
(0.0158)
(0.0142)
(0.0187)
(0.0189)
(0.0168)
(0.0140)
(0.0155)
(0.0206)
(0.0227)

Below

Poverty

-0.0062

(0.0044)

-9.5E-06 (5.9E-05)

0.1331
0.1071
0.0908
0.0148
0.2693

-0.2086
2.7E-05
0.9581
0.1364
0.3289
-0.0235
0.1652
0.0096
0.0226
0.0457
0.0278
0.0323
0.0292
0.0589
0.0915
0.0685
0.1003

41,708
-20,698

Estimates are partial derivatives evaluated at the sample means.

194

(0.0114)
(0.0042)
(0.0065)
(0.0131)
(0.0066)

(0.0553)

(2.51305)

(0.3169)
(0.1827)
(0.2294)
(0.0173)
(0.1820)
(0.0146)
(0.0196)
(0.0178)
(0.0221)
(0.0221)
(0.0193)
(0.0165)
(0.0180)
(0.0230)
(0.0249)

Below 150%
of Poverty Level

-0.0071

(0.0048)

0.0000 (6.5E-05)

0.1582
0.1362
0.0925
0.0041
0.3008

-0.1561
2.7E-05
1.0185
0.3620
0.1532
-0.0169
0.3927
0.0084
0.0230
0.0376
0.0343
0.0441
0.0242
0.0554
0.0817
0.0870
0.11 17

41,708
-23,017

(0.0113)
(0.0056)
(0.0083)
(0.0144)
(0.0066)

(0.0062)

(2.8E-05)

(0.3493)
(0.2045)
(0.2511)
(0.0199)
(0.1968)
(0.0161)
(0.0210)
(0.0188)
(0.0236)
(0.0235)
(0.0207)
(0.0171)
(0.0182)
(0.0232)
(0.0245)

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Non labor income
Federal subsidy rate
State subsidy rate
AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. All regressions include full set of state dummies.

TABLE A-lO
Probits Using Separate Federal and State Subsidy Rates with
Low-Education Subsample

Labor-force
Participation

0.0176
-0.0002
-0. 1099
-0.0366
-0.0754

0.0095
-0. 1976
-0.0057

0.2495
-0.0673
1.6E-04
-1.3868

0.2436
-O.3487
-0.0012
-0.1699
-0.001 1
-0.0152
-0.0245
-0.0386
-0.0487
-0.0488
-0.0595
-0.0747
-0.0787
-0.1010

41,708
-l7,014

(0.0034)

(4.6E-05)

(0.0100)
(0.0029)
(0.0046)
(0.0098)
(0.0057)
(0.0005)
(0.0483)
(0.2127)

(2.0805)

(0.2537)
(0.1465)
(0.1857)
(0.0136)
(0.1396)
(0.0113)
(0.0157)
(0.0142)
(0.0186)
(0.0189)
(0.0168)
(0.0140)
(0.0155)
(0.0206)
(0.0227)

Below

Poverty

-0.0062
-9.6E-06
0.1330
0.1072
0.0908
0.0149
0.2693

-0.2227
-0.0403
2.7E-05
0.9364
0.1562
0.3297
-0.0214
0.1658
0.0096
0.0223
0.0455
0.0276
0.0321
0.0289
0.0590
0.0914
0.0688
0.1004

41 ,708
-20,698

Estimates are partial derivatives evaluated at the sample means.

195

(0.0044)

(5.9E-05)

(0.0114)
(0.0042)
(0.0065)
(0.0131)
(0.0066)

(0.0595)
(0.3673)

(2.51305)

(0.3184)
(0.1840)
(0.2293)
(0.0180)
(0.1820)
(0.0146)
(0.0196)
(0.0178)
(0.0221)
(0.0221)
(0.0193)
(0.0165)
(0.0180)
(0.0230)
(0.0249)

Below 150%
of Poverty Level

-0.007l
~2.4E-06
0.1579
0.1365
0.0925
0.0043
0.3008

-0.1975
0.3584
2.7E-05
0.9577
0.4207
0.1564
-0.0106
0.3954
0.0083
0.0220
0.0369
0.0337
0.0433
0.0234
0.0554
0.0814
0.0877
0.1116

41,708
-23,015

(0.0048)

(6515-05)

(0.01 13)
(0.0056)
(0.0084)
(0.0144)
(0.0066)

(0.0668)
(0.4673)

(2.8E-05)

(0.3503)
(0.2054)
(0.2510)
(0.0208)
(0.1967)
(0.0161)
(0.0210)
(0.0188)
(0.0236)
(0.0235)
(0.0208)
(0.0171)
(0.0182)
(0.0232)
(0.0245)

Probits Using Separate Federal and Refundable State Rates with

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Non labor income
Federal subsidy rate
Refundable state EITC
AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. A11 regressions include full set of state dummies.

TABLE A-ll

Low-Education Subsample

Labor-force
Participation

0.0176
-0.0002
-0.1099
-0.0366
-0.0754

0.0094
-0.l977
-0.0057

0.2452

0.0469
1.6E-04
-1.3979

0.2514
-0.3506
-0.0003
-0.1693
-0.0010
-0.0153
-0.0246
-0.0387
-0.0489
-0.0491
-0.0596
-0.0749
-0.0791
-0. 1017

41,708
-17,015

(0.0034)

(4.6E-05)

(0.0100)
(0.0029)
(0.0046)
(0.0098)
(0.0057)
(0.0005)
(0.0483)
(0.3073)

(2.0505)

(0.2534)
(0.1462)
(0.1858)
(0.0137)
(0.1396)
(0.0113)
(0.0157)
(0.0142)
(0.0187)
(0.0189)
(0.0168)
(0.0140)
(0.0155)
(0.0206)
(0.0227)

Below

Poverty

-0.0062

(0.0044)

—9.7E-06 (5.9E-05)

0.1329
0.1071
0.0908
0.0150
0.2693

-0.2289
0.1761
2.7E-05
0.9222
0.1682
0.3256
-0.0199
0.1654
0.0097
0.0224
0.0454
0.0275
0.0319
0.0286
0.05 89
0.091 1
0.0685
0.0996

41 ,708
-20,698

Estimates are partial derivatives evaluated at the sample means.

196

(0.0114)
(0.0042)
(0.0065)
(0.0131)
(0.0066)

(0.0591)
(0.4373)

(2.5E-05)

(0.3183)
(0.1844)
(0.2295)
(0.0178)
(0.1819)
(0.0146)
(0.0196)
(0.0178)
(0.0221)
(0.0221)
(0.0193)
(0.0165)
(0.0180)
(0.0229)
(0.0249)

Below 150%
of Povem Level

-0.0071
—2.7E—06
0.1578
0.1365
0.0926
0.0048
0.3008

-0.1884
0.2012
2.8E-05
0.9822
0.4011
0.1516
-0.0116
0.3913
0.0084
0.0227
0.0374
0.0342
0.0439
0.0240
0.0560
0.0819
0.0885
0.1126

41,708
-23,016

(0.0048)

(6.5E-05)

(0.01 13)
(0.0056)
(0.0083)
(0.0144)
(0.0066)

(0.0658)
(0.5006)

(2.8E-05)

(0.3502)
(0.2064)
(0.2511)
(0.0205)
(0.1967)
(0.0161)
(0.0210)
(0.0188)
(0.0236)
(0.0235)
(0.0207)
(0.0171)
(0.0182)
(0.0232)
(0.0245)

TABLE A-12

Probits Using Combined Federal and State Subsidy Rate with
Year/Child Interactions

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Some college

Degree

Low education x Kids
Nonwhite x low education
Non labor income
Subsidy rate

AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Kids x Year - 1983
Kids x Year - 1984
Kids x Year - 1985
Kids x Year - 1988
Kids x Year - 1989
Kids x Year - 1990
Kids x Year - 1992
Kids x Year - 1993
Kids x Year - 1995
Kids x Year - 1996
Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. A11 regressions include full set of state dummies.

Labor-force
Participation

0.0133
-0.0002
-0.0684
-0.0251
-0.0499

0.0136
-0. 1532

0.0405

0.0406

0.0031
-0.01 12
-0.0029

0.0678

-9.1E-05
-1.0963

0.1649
-0.2492

0.0092
-0.1 173
-0.0051

0.0108

0.0042
-0.0123

0.0144

0.0152
-0.0027

0.0008

0.0102

0.0151

0.0047
-0.0156
—0.0196
—0.0179
-0.0452
-0.0412
-0.0247
-0.0440
-0.0512
-0.0712

83,935
-25,759

(0.0018)

(2,413-05)

(0.0066)
(0.0017)
(0.0025)
(0.0048)
(0.0051)
(0.0044)
(0.0034)
(0.0049)
(0.0057)
(0.0001)
(0.0897)

(1.2505)

(0.1405)
(0.1047)
(0.0952)
(0.0073)
(0.0723)
(0.01 10)
(0.0090)
(0.0098)
(0.0135)
(0.0105)
(0.0106)
(0.0141)
(0.0148)
(0.0270)
(0.0281)
(0.0086)
(0.0109)
(0.0104)
(0.0125)
(0.0135)
(0.0123)
(0.0096)
(0.0109)
(0.0130)
(0.0141)

Below

Poverg

-0.0085
5.8E-05
0.0578
0.0708
0.0698
0.0166
0.2243
-0.0510
-0.0858
0.0112
0.0372

-0.0086
-6.1E-07
0.7993
-0.0948
0.1871
-0.0183
0.0317
0.0149
0.0120
0.0332
-0.0064
-0.0197
—0.0193
-0.0050
—0.0171
-0.0247
-0.0224
-0.0091
0.0005
-0.0043
0.0022
0.0108
0.0110
0.0155
0.0398
0.0256
0.0460

83,935
-32,559

Estimates are partial derivatives evaluated at the sample means.

197

(0.0024)

(3.3505)

(0.0078)
(0.0025)
(0.0038)
(0.0076)
(0.0062)
(0.0055)
(0.0042)
(0.0067)
(0.0081)

(0.1371)

(1.6E-05)

(0.1842)
(0.1436)
(0.1235)
(0.0101)
(0.0990)
(0.0149)
(0.0141)
(0.0156)
(0.0170)
(0.0158)
(0.0156)
(0.0198)
(0.0204)
(0.0399)
(0.0433)
(0.0112)
(0.0131)
(0.0120)
(0.0144)
(0.0145)
(0.0131)
(0.0114)
(0.0126)
(0.0140)
(0.0151)

Below 150%
of Poverty Lag!

-0.0136
0.0001
0.0865
0.1054
0.0857
0.0198
0.2860

-0.0834

-0.l343
0.0195
0.0425

-0.0685
-2.7E-05
0.9385
0.1500
0.2492
-0.0075
0.2253
0.0326
0.0287
0.0424
0.0264
-0.0065
-0.0051
0.01 19
0.0031
0.0007
0.0054
-0.0256
-0.0132
-0.01 19
-0.0105
0.0073
-0.0024
0.0016
0.0340
0.0358
0.0495

83,935
-39,3 86

(0.0030)

(4.1E-05)

(0.0091)
(0.0037)
(0.0054)
(0.0094)
(0.0069)
(0.0065)
(0.0051)
(0.0081)
(0.0097)

(0.1838)

(2.0505)

(0.2287)
(0.1836)
(0.1533)
(0.0131)
(0.1211)
(0.0187)
(0.0178)
(0.0188)
(0.0233)
(0.0217)
(0.0216)
(0.0270)
(0.0289)
(0.0588)
(0.0634)
(0.0132)
(0.0155)
(0.0144)
(0.0172)
(0.0172)
(0.0154)
(0.0132)
(0.0143)
(0.0166)
(0.0174)

TABLE A-l3

Probits Using Separate Federal and State Subsidy Rates with
Year/Child Interactions

Labor-force
ELM

Age 0.0133
Age squared -0.0002
Nonwhite -0.0684
Number of children -0.0262
Child under 6 -0.0500
Nonwhite x kids 0.0134
< High school 01530
Some college 0.0400
Degree 0.0409
Low education x Kids 0.0018
Nonwhite x low education -0.0110
Non labor income -0.0029
Federal subsidy rate 0.3409
State subsidy rate -0.1340
AFDC maximum x kids -9.5E-05
Unemployment rate -0.9797
Unemployment x kids -0.0101
Personal income -0.2432
Welfare waiver x kids 0.0081
Minimum wage -0.0121
Kids x Year - 1983 -0.0103
Kids x Year - 1984 0.0028
Kids x Year - 1985 -0.0078
Kids x Year - 1988 -0.0395
Kids x Year - 1989 -0.0083
Kids x Year - 1990 -0.0066
Kids x Year - 1992 -0.0370
Kids x Year - 1993 -0.0372
Kids x Year - 1995 -0.0864
Kids x Year - 1996 -0.0873
Year dummy - 1983 0.0072
Year dummy - 1984 -0.0103
Year dummy - 1985 -0.0135
Year dummy - 1988 -0.0098
Year dummy - 1989 -0.0355
Year dummy - 1990 -0.0319
Year dummy - 1992 00189
Year dummy - 1993 -0.0367
Year dummy - 1995 00419
Year dummy - 1996 -0.0608
Observations 83,935
Log Likelihood -25,754

Standard errors in parentheses. All regressions include full set of state dummies.

(0.0018)

(2.4E-05)

(0.0066)
(0.0017)
(0.0026)
(0.0048)
(0.0051)
(0.0044)
(0.0033)
(0.0050)
(0.0057)
(0.0001)
(0.1348)
(0.1148)

(1.2805)

(0.1470)
(0.1229)
(0.0953)
(0.0073)
(0.0723)
(0.0115)
(0.0100)
(0.0115)
(0.0190)
(0.0154)
(0.0152)
(0.0221)
(0.0241)
(0.0633)
(0.0681)
(0.0085)
(0.0106)
(0.0102)
(0.0121 )
(0.0133)
(0.0121 )
(0.0095)
(0.0107)
(0.0128)
(0.0139)

Below
Poverty

-0.0085
5.8E-05
0.0578
0.0712
0.0699
0.0167
0.2243
-0.0509
-0.0859
0.0115
0.0371

-0.0899
0.0529
4.7E-07
0.7667
-0.0424
0.1859
-0.0180
0.0332
0.0164
0.0147
0.0371
0.0004
-0.0133
-0.0130
0.0039
-0.0075
-0.0025
0.0018
-0.0098
-0.0009
-0.0058
0.0001
0.0086
0.0088
0.0140
0.0380
0.0234
0.0436

83,935
-32,559

Estimates are partial derivatives evaluated at the sample means.

198

(0.0024)

(3.3805)

(0.0078)
(0.0026)
(0.0038)
(0.0076)
(0.0062)
(0.0055)
(0.0042)
(0.0067)
(0.0081)

(0.1773)
(0.1980)

(1.6E-05)

(0.1904)
(0.1609)
(0.1236)
(0.0101)
(0.0990)
(0.0151)
(0.0147)
(0.0166)
(0.0197)
(0.0185)
(0.0181)
(0.0239)
(0.0249)
(0.0536)
(0.0587)
(0.0112)
(0.0131 )
(0.0120)
(0.0146)
(0.0147)
(0.0133)
(0.0115)
(0.0128)
(0.0142)
(0.0153)

Below 150%
of Povert Level

-0.0136
0.0001
0.0865
0.1063
0.0858
0.0199
0.2860

-0.0832

-0. 1344
0.0201
0.0424

-0.2247
0.0503
-2.5E-05
0.8813
0.251 1
0.2472
-0.0070
0.2278
0.0356
0.0338
0.0496
0.0402
0.0067
0.0079
0.0297
0.0228
0.0473
0.0556
-0.0269
-0.0155
0.0145
-0.0140
0.0036
-0.0060
-0.0009
0.0310
0.0321
0.0456

83,935
-39,386

(0.0030)

(4.1805)

(0.0091)
(0.0038)
(0.0054)
(0.0094)
(0.0069)
(0.0065)
(0.0051)
(0.0082)
(0.0097)

(0.2255)
(0.2696)

(2.0805)

(0.2346)
(0.2020)
(0.1534)
(0.0131)
(0.121 1)
(0.0189)
(0.0184)
(0.0199)
(0.0264)
(0.0248)
(0.0244)
(0.0312)
(0.0337)
(0.0736)
(0.0799)
(0.0132)
(0.0155)
(0.0145)
(0.0174)
(0.0174)
(0.0156)
(0.0133)
(0.0145)
(0.0168)
(0.0176)

TABLE A-l4

Probits Using Separate Federal and Refundable State Rates with
Year/Child Interactions

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Some college

Degree

Low education x Kids
Nonwhite x low education
Non labor income
Federal subsidy rate
Refundable state EITC
AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage

Kids x Year - 1983
Kids x Year - 1984
Kids x Year - 1985
Kids x Year - 1988
Kids x Year - 1989
Kids x Year - 1990
Kids x Year - 1992
Kids x Year - 1993
Kids x Year - 1995
Kids x Year - 1996
Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. All regressions include full set of state dummies.

Labor-force
28mm

0.0133
-0.0002
-0.0683
-0.0262
-0.0500

0.0132
-0.1531

0.0400

0.0409

0.0018
-0.0109
-0.0029

0.3374
-0.0367

-9.6E-05
-0.9925
-0.0007
-0.2439

0.0091
-0.1 196
-0.0103

0.0029
-0.0076
-0.0390
-0.0080
-0.0063
-0.0370
-0.0374
-0.0868
-0.0883

0.0073
-0.0105
-0.0138
-0.0102
-0.0361
-0.0325
-0.0192
-0.0371
-0.0424
—0.0614

83,935
-25,755

(0.0018)

(2.4E-05)

(0.0066)
(0.0017)
(0.0026)
(0.0048)
(0.0051)

(0.0044) _

(0.0033)
(0.0050)
(0.0057)
(0.0001)
(0.1348)
(0.1730)

(1.2505)

(0.1469)
(0.1228)
(0.0953)
(0.0074)
(0.0723)
(0.0115)
(0.0100)
(0.0115)
(0.0190)
(0.0154)
(0.0151)
(0.0221)
(0.0242)
(0.0634)
(0.0684)
(0.0085)
(0.0107)
(0.0102)
(0.0122)
(0.0133)
(0.0121)
(0.0095)
(0.0107)
(0.0128)
(0.0139)

Below
Poverty

-0.0085
5.8E-05
0.0577
0.0712
0.0699
0.0168
0.2243
-0.0509
-0.0859
0.0116
0.0371

-0.0905
0.1531
-1 .8E-08
0.7636
-0.0396
0.1826
-0.0172
0.0322
0.0165
0.0149
0.0374
0.0007
-0.0131
-0.0127
0.0040
-0.0076
-0.0028
0.0012
-0.0098
-0.0008
-0.0058
0.0000
0.0085
0.0086
0.0139
0.0379
0.0233
0.0435

83,935
-32,559

Estimates are partial derivatives evaluated at the sample means.

199

(0.0024)

(3.31505)

(0.0078)
(0.0026)
(0.0038)
(0.0076)
(0.0062)
(0.0055)
(0.0042)
(0.0067)
(0.0081)

(0.1775)
(0.2458)

(1.6E-05)

(0.1905)
(0.1614)
(0.1236)
(0.0101)
(0.0990)
(0.0151)
(0.0147)
(0.0167)
(0.0197)
(0.0186)
(0.0181)
(0.0239)
(0.0249)
(0.0535)
(0.0586)
(0.0112)
(0.0131)
(0.0120)
(0.0146)
(0.0147)
(0.0133)
(0.01 15)
(0.0128)
(0.0142)
(0.0153)

Below 150%
of Poverty Level

-0.0136
0.0001
0.0865
0.1063
0.0858
0.0200
0.2860

-0.0831

-0. 1344
0.0202
0.0424

—0.2252
0.1411
-2.5E-05
0.8790
0.2536
0.2444
-0.0063
0.2271
0.0356
0.0340
0.0498
0.0405
0.0070
0.0081
0.0298
0.0228
0.0471
0.0551
-0.0268
—0.0154
-0.0145
-0.0141
0.0035
-0.0061
—0.0009
0.0309
0.0320
0.0455

83,935
-39,3 86

(0.0030)

(4.11505)

(0.0091)
(0.0038)
(0.0054)
(0.0094)
(0.0069)
(0.0065)
(0.0051)
(0.0082)
(0.0097)

(0.2258)
(0.3344)

(2.0E-05)

(0.2349)
(0.2031)
(0.1534)
(0.0130)
(0.121 1)
(0.0189)
(0.0184)
(0.0199)
(0.0264)
(0.0249)
(0.0245)
(0.0312)
(0.0337)
(0.0735)
(0.0798)
(0.0132)
(0.0155)
(0.0145)
(0.0174)
(0.0174)
(0.0156)
(0.0133)
(0.0145)
(0.0168)
(0.0176)

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Non labor income
Subsidy rate

AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage
Kids x Year - 1983
Kids x Year - 1984
Kids x Year - 1985
Kids x Year - 1988
Kids x Year — 1989
Kids x Year - 1990
Kids x Year - 1992
Kids x Year - 1993
Kids x Year - 1995
Kids x Year - 1996
Year dummy - 1983
Year dummy - 1984
Year dummy — 1985
Year dummy - 1988
Year dunnny - 1989
Year dummy - 1990
Year durmny - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. All regressions include full set of state dtunmies.

TABLE A-lS
Probits Using Combined Federal and State Subsidy Rate with
Year/Child Interactions and Low-Education Subsample

Labor-force
Participation

0.0176
-0.0002
-0. 1093
-0.0363
-0.0755

0.0084
-0.1975
-0.0057

0.1014
1.6E-04
-1.4790

0.3747
-0.3487

0.0031
-0. 1691
-0.0004

0.0169
-0.0220
-0.0061

0.0281

0.0412

0.0253

0.0174

0.0360

0.0385
-0.0009
-0.0275
-0.0093
-0.0306
-0.0668
-0.0778
-0.0713
-0.0800
-0.0834
—0.1063

41,708
-17,007

(0.0034)

(4.6E-05)

(0.0100)
(0.0030)
(0.0046)
(0.0099)
(0.0057)
(0.0005)
(0.1636)

(2.31805)

(0.2642)
(0.1865)
(0.1855)
(0.0139)
(0.1392)
(0.0195)
(0.0171)
(0.0205)
(0.0234)
(0.0198)
(0.0186)
(0.0235)
(0.0265)
(0.0467)
(0.0502)
(0.0173)
(0.0209)
(0.0186)
(0.0232)
(0.0242)
(0.0226)
(0.0199)
(0.0213)
(0.0246)
(0.0262)

Below

Povem

-0.0062

(0.0044)

-9.3E-06 (5.9E-05)

0.1321
0.1059
0.0915
0.0163
0.2693

0.0742
2213-05
1.1 146
-0.1502
0.3290
-0.0210
0.1599
0.0124
-0.0024
0.0729
-0.0096
-0.0487
-0.0483
-0.0441
-0.0493
-0.0764
-0.0771
0.0017
0.0231
-0.0013
0.0253
0.0551
0.0516
0.0717
0.1058
0.0719
0.0992

41 ,708
-20,680

Estimates are partial derivatives evaluated at the sample means.

200

(0.0114)
(0.0043)
(0.0065)
(0.0132)
(0.0066)

(0.2581)

(2.9E-05)

(0.3340)
(0.2565)
(0.2291)
(0.0184)
(0.1814)
(0.0253)
(0.0239)
(0.0273)
(0.0313)
(0.0288)
(0.0284)
(0.0353)
(0.0380)
(0.0726)
(0.0774)
(0.0213)
(0.0250)
(0.0226)
(0.0275)
(0.0275)
(0.0249)
(0.0227)
(0.0241)
(0.0271)
(0.0284)

Below 150%
of Pover_ty Level

-0.0073

(0.0048)

0.0000 (6.5E-05)

0.1580
0.1336
0.0935
0.0047
0.3009

0.2512
8.8E-06
1.2535
-0. 1038
0.1580
-0.0118
0.3863
0.0485
0.0205
0.0774
0.0060
-0.0452
-0.0559
-0.0631
-0.0625
-0. 1091
-0.1059
-0.0191
0.0097
-0.0092
0.0198
0.0575
0.0436
0.0693
0.0923
0.0871
0.1043

41 ,708
-22,995

(0.01 13)
(0.0058)
(0.0084)
(0.0145)
(0.0066)

(0.3171)

(3.2E-05)

(0.3681)
(0.2968)
(0.251 1)
(0.0214)
(0.1961)
(0.0274)
(0.0265)
(0.0280)
(0.0365)
(0.0356)
(0.0350)
(0.0436)
(0.0479)
(0.0954)
(0.1014)
(0.0219)
(0.0257)
(0.0238)
(0.0285)
(0.0278)
(0.0253)
(0.0222)
(0.0231)
(0.0268)
(0.0277)

Age

Age squared
Nonwhite

Number of children
Child under 6
Nonwhite x kids

< High school

Non labor income
Federal subsidy rate
State subsidy rate
AFDC maximum x kids
Unemployment rate
Unemployment x kids
Personal income
Welfare waiver x kids
Minimum wage
Kids x Year - 1983
Kids x Year - 1984
Kids x Year - 1985
Kids x Year - 1988
Kids x Year - 1989
Kids x Year - 1990
Kids x Year - 1992
Kids x Year - 1993
Kids x Year - 1995
Kids x Year - 1996
Year dummy - 1983
Year dummy - 1984
Year dummy - 1985
Year dummy - 1988
Year dummy - 1989
Year dummy - 1990
Year dummy - 1992
Year dummy - 1993
Year dummy - 1995
Year dummy - 1996

Observations
Log Likelihood

Standard errors in parentheses. All regressions include full set of state dummies.

TABLE A-16
Probits Using Separate Federal and State Subsidy Rates with
Year/Child Interactions and Low-Education Subsample

Labor-force
Min—ting

0.0176
-0.0002
-0.1089
-0.0371
-0.0755

0.0080
-0. 1974
-0.0057

0.3276
-0.0561
1.6E-04
- l .3689

0.2229
-0.3454

0.0022
-0. 1718
-0.0043

0.0102
-0.0328
-0.0264

0.0099

0.0246

0.0014
-0.0104
-0.0282
-0.031 1

0.0015
-0.0225
-0.0040
-0.0229
-0.0581
-0.0692
-0.0654
-0.0733
-0.0749
-0.0970

41,708
-l7,007

(0.0034)

(4.6E-05)

(0.0100)
(0.0031)
(0.0046)
(0.0099)
(0.0057)
(0.0005)
(0.2558)
(0.2133)

(2.3E-05)

(0.2806)
(0.2280)
(0.1855)
(0.0140)
(0.1393)
(0.0199)
(0.0183)
(0.0230)
(0.0307)
(0.0266)
(0.0248)
(0.0333)
(0.0381)
(0.0858)
(0.0931)
(0.0172)
(0.0208)
(0.0186)
(0.0235)
(0.0246)
(0.0231)
(0.0201)
(0.0215)
(0.0250)
(0.0266)

Below

Poverty

-0.0062
-9.2E-06
0.1323
0.1051
0.0915
0.0160
0.2694

0.25 89
-0.05 83
1.9E-05

1.1983
-0.2742

0.3296
-0.0216

0.1570

0.0091
-0.0081

0.0640
-0.0249
-0.0631
-0.0624
-0.0628
-0.0700
-0. 1201
-0. 1240

0.0037

0.0269

0.0027

0.0309

0.061 1

0.0574

0.0759

0.1 106

0.0777

0.1054

41,708
-20,680

Estimates are partial derivatives evaluated at the sample means.

201

(0.0044)

(595.05)

(0.01 14)
(0.0044)
(0.0065)
(0.0132)
(0.0066)

(0.3240)
(0.3706)

(2.91305)

(0.3476)
(0.2873)
(0.2292)
(0.0185)
(0.1815)
(0.0254)
(0.0245)
(0.0285)
(0.0342)
(0.0316)
(0.0309)
(0.0387)
(0.0415)
(0.0762)
(0.0813)
(0.0215)
(0.0254)
(0.0230)
(0.0282)
(0.0284)
(0.0257)
(0.0232)
(0.0246)
(0.0280)
(0.0293)

Below 150%
of Povert Level

-0.0073

(0.0048)

0.0000 (6.5E-05)

0.1579
0.1343
0.0935
0.0048
0.3009

0.1265
0.3444
1.1E-05
1.2013
-0.0200
0.1577
-0.0114
0.3883
0.0507
0.0244
0.0828
0.0166
—0.0347
-0.0457
-0.0497
-0.0474
-0.0749
-0.0692
-0.0203
0.0075
-0.01 16
0.0165
0.0540
0.0403
0.0670
0.0897
0.0838
0.1009

41 ,708
-22,995

(0.0113)
(0.0059)
(0.0084)
(0.0145)
(0.0066)

(0.3590)
(0.4718)

(3.2E-05)

(0.3776)
(0.3176)
(0.251 1)
(0.0216)
(0.1961)
(0.0275)
(0.0269)
(0.0288)
(0.0390)
(0.0381)
(0.0373)
(0.0465)
(0.051 1)
(0.1059)
(0.1 142)
(0.0220)
(0.0258)
(0.0240)
(0.0288)
(0.0282)
(0.0257)
(0.0223)
(0.0234)
(0.0272)
(0.0281)

TABLE A-l7
Probits Using Separate Federal and Refundable State Rates with
Year/Child Interactions and Low-Education Subsample

Labor-force Below Below 150%
Participation Poverty of Poverty Level

Age 0.0176 (0.0034) -0.0062 (0.0044) -0.0073 (0.0048)
Age squared -0.0002 (4.6E-05) -9.4E-06 (5.9E-05) 0.0000 (6.5E-05)
Nonwhite -0.1089 (0.0100) 0.1323 (0.01 14) 0.1577 (0.0113)
Number of children —0.0372 (0.0031) 0.1050 (0.0044) 0.1343 (0.0059)
Child under 6 -0.0755 (0.0046) 0.0915 (0.0065) 0.0935 (0.0084)
Nonwhite x kids 0.0080 (0.0099) 0.0160 (0.0132) 0.0054 (0.0145)
< High school -0. 1975 (0.0057) 0.2693 (0.0066) 0.3009 (0.0066)
Non labor income -0.0057 (0.0005)

Federal subsidy rate 0.3258 (0.2557) 0.2549 (0.3245) 0.1353 (0.3600)
State subsidy rate 0.0748 (0.3097) 0.1631 (0.4394) 0.1909 (0.5059)
AFDC maximum x kids -0.0002 (2.3E-05) 1.8E-05 (2.9E-05) 1.2E-05 (3.2E-05)
Unemployment rate -1.3796 (0.2802) 1.1839 (0.3480) 1.2257 (0.3786)
Unemployment x kids 0.2296 (0.2277) -0.2627 (0.2885) -0.0405 (0.3201)
Personal income -0.3479 (0.1856) 0.3262 (0.2294) 0.1529 (0.2512)
Welfare waiver x kids 0.0033 (0.0141) -0.0200 (0.0184) -0.0127 (0.0213)
Minimum wage -0. 1717 (0.1393) 0.1566 (0.1815) 0.3846 (0.1962)
Kids x Year - 1983 -0.0043 (0.0199) 0.0091 (0.0254) 0.0507 (0.0275)
Kids x Year - 1984 0.0103 (0.0183) -0.0079 (0.0245) 0.0244 (0.0269)
Kids x Year - 1985 -0.0326 (0.0230) 0.0644 (0.0285) 0.0827 (0.0288)
Kids x Year - 1988 -0.0261 (0.0307) -0.0243 (0.0343) 0.0159 (0.0391)
Kids x Year - 1989 0.0101 (0.0266) -0.0627 (0.0316) -0.0353 (0.0382)
Kids x Year - 1990 0.0247 (0.0248) -0.0620 (0.0310) -0.0463 (0.0373)
Kids x Year - 1992 0.0014 (0.0333) -0.0628 (0.0387) -0.0498 (0.0465)
Kids x Year - 1993 -0.0106 (0.0382) -0.0701 (0.0415) -0.0475 (0.0510)
Kids x Year - 1995 -0.0289 (0.0859) -0. 1205 (0.0761) -0.0751 (0.1058)
Kids x Year - 1996 -0.0322 (0.0935) -0. 1248 (0.0811) -0.0688 (0.1141)
Year dummy - 1983 0.0016 (0.0172) 0.0038 (0.0215) -0.0202 (0.0220)
Year dtunmy - 1984 -0.0226 (0.0208) 0.0268 (0.0254) 0.0081 (0.0258)
Year dummy - 1985 -0.0042 (0.0186) 0.0024 (0.0230) -0.0111 (0.0240)
Year dummy - 1988 -0.0234 (0.0235) 0.0303 (0.0282) 0.0174 (0.0289)
Year dummy - 1989 -0.0586 (0.0247) 0.0604 (0.0284) 0.0550 (0.0282)
Year dummy - 1990 -0.0697 (0.0232) 0.0567 (0.0258) 0.0412 (0.0257)
Year dummy - 1992 -0.0656 (0.0201) 0.0756 (0.0232) 0.0676 (0.0224)
Year dummy - 1993 -0.0737 (0.0215) 0.1101 (0.0246) 0.0903 (0.0234)
Year dummy - 1995 -0.0754 (0.0250) 0.0771 (0.0280) 0.0847 (0.0272)
Year dummy - 1996 -0.0976 (0.0266) 0.1047 (0.0293) 0.1018 (0.0282)
Observations 41 ,708 41,708 41 ,708

Log Likelihood -17,007 -20,680 -22,996

Standard errors in parentheses. All regressions include full set of state dummies.

Estimates are partial derivatives evaluated at the sample means.

202

BIBLIOGRAPHY

203

BIBLIOGRAPHY

Agell, Jonas, and Kjell Erik Lommerud. “Minimum Wages and the Incentives for Skill
Formation.” Journal of Public Economics, Vol. 64, (1997): pp. 25 - 40.

Alstott, Anne L. "The Earned Income Tax Credit and Some Fundamental Institiutional
Dilemmas of Tax-Transfer Integration." National Tax Journal, Vol. XLVH, No.3,
(Sept. 1994): pp. 609 - 619.

Barron, John M., Mark C. Berger, and Dan A. Black. “Do Workers Pay for On-the Job
Training?” Journal of Human Resources, Vol. 34, (Spring 1999a): pp. 235 - 252.

Barron, John M., Mark C. Berger, and Dan A. Black. “Replacing General with Speciﬁc
Training: Why Restricting Alternatives Makes Sense.” Research in Labor
Economics, Vol. 18, (1999b): pp. 281 - 302.

Barrow, Lisa, and Leslie McGranahan. “The Effects of the Earned Income Credit on the
Seasonality of Household Expenditures.” National Tax Journal, Vol. LIII, No. 4,
Part 2, (Dec. 2000): pp. 1211 - 1243.

Becker, Gary S. Human Capital. National Bureau of Economic Research and University
of Chicago Press, Chicago, IL: 1993.

Blank, Rebecca M. and Patricia Ruggles. “When Do Women Use AFDC and Food
Stamps? The Dynamics of Eligibility Versus Participation.” Mimeo.
Washington, DC: Northwestern University and the Urban Institute, 1993.

Brown, Charles, Curtis Gilroy, and Andrew Kohen. “The Effect of the Minimum Wage
on Employment and Unemployment.” Journal of Economic Literature, Vol. XX,
(June 1982): pp. 487 - 528.

Browning, Edgar K. "Effects of the Earned Income Tax Credit on Income and Welfare."
National Tax Journal, Vol. XLVHI, No. 3, (March 1995): pp. 23 - 43.

Browning, Edgar K. and J acquelene M. Browning. Public Finance and the Price
System. New Jersey: Prentice Hall, 1994.

Burkhauser, Richard V., Kenneth A. Couch, and Andrew J. Glenn. “Public Policies for
the Working Poor.” Research in Labor Economics, Volume 15, 1996: pp. 65 -
109.

Burkhauser, Richard V., Kenneth A. Couch, and David C. Wittenburg. “‘Who Gets
What’ from Minimum Wage Hikes: A Re-Estimation of Card and Krueger’s
Distributional Analysis in Myth and Measurement: The New Economics of the
Minimum Wage.” Industrial and Labor Relations Review, Vol. 49, No.3, April

204

1996: pp. 547 - 552.

Chemick, Howard and Andrew Reschovsky. "The Taxation of the Poor." Journal of
Human Resources, Vol. XXV, No. 4, (Fall 1990): pp. 712 - 735.

Council of Economic Advisers. “The Effects of Welfare Policy and the Economic
Expansion on Welfare Caseloads: An Update.” Technical Report. Washington,
DC: U.S. Government Printing Ofﬁce, August 1999.

Council of Economic Advisers. “Explaining the Decline in Welfare Receipt, 1993-
1996.” Technical Report. Washington, DC: U.S. Government Printing Ofﬁce,
May 1997.

Council of Economic Advisers. “Good News for Low Income Families: Expansions in
the Earned Income Credit and the Minimum Wage.” Technical Report.
Washington, DC: U.S. Government Printing Ofﬁce, May 1997.

Cunningham, James. “The Impact of Minimum Wages on Youth Employment, Hours of
Work, and School Attendance: Cross-sectional Evidence from the 1960 and 1970
Censuses.” In The Economics of Legal Minimum Wages, edited by Simon
Rottenberg. Washington, DC: American Enterprise Institute for Public Policy
Research, 1981, pp. 88 - 123.

Dickert, Stacy, Scott Houser, and John Karl Scholz. “The Earned Income Tax Credit and
Transfer Programs: A Study of Labor Market and Program Participation.” In Tax
Policy and the Economy, Vol. 9, edited by James Poterba. Cambridge, MA:
NBER and MIT Press Journals, 1995, pp. 1 - 50.

Dickert, Stacy, Scott Houser, and John Karl Scholz. "Taxes and the Poor: A
Microsimulation Study of Implicit and Explicit Taxes." National Tax Journal,
Vol. XLVII, No. 3, (Sept. 1994): 621 - 638.

Ehrenberg, Ronald G., and Alan J. Marcus. “Minimum Wages and Teenagers’
Enrollment-Employment Outcomes: A Multinomial Logit Model.” The Journal
of Human Resources, XVII, (1982): 39 - 52.

Ehrenberg, Ronald G., and Alan J. Marcus. “Minimum Wage Legislation and the
Educational Outcomes of Youths.” Research in Labor Economics, Volume 3,
1980, pp. 61 - 93.

Eissa, Nada. "Tax Reforms and Labor Supply." In Tax Policy and the Economy, Vol. 10,
edited by James Poterba. Cambridge, MA: NBER and MIT Press Journals, 1996,
119 - 151.

Eissa, Nada. "Taxation and Labor Supply of Married Women: The Tax Reform Act of

205

1986 as Natural Experiment." Working Paper No. 5023. Cambridge, MA:
NBER, 1995.

Eissa, Nada and Jeffrey B. Liebman. "Labor Supply Response to the Earned Income Tax
Credit." Quarterly Journal of Economics, Vol. 111, No. 2. (May 1996): 605 -
637.

Eissa, Nada and Hilary Williamson Hoynes. "The Earned Income Tax Credit and the
Labor Supply of Married Couples." Working Paper No. 4. Berkeley, CA: Center
for Labor Economics, 1998.

Ellwood, David T. “The Impact of the Earned Income Tax Credit and Social Policy
Reforms on Work, Marriage, and Living Arrangements.” National Tax Journal,
Vol. LIII, No.4, Part 2, (Dec. 2000): 1063 - 1106.

Fullerton Jr., Howard N. “Labor Force Participation: 75 years of change, 1950-98 and
1998-2025.” Monthly Labor Review. U.S. Department of Labor, Dec. 1999: pp.
3 - 12.

Hashimoto, Masanori. “Minimum Wage Effects on Training on the Job.” American
Economic Review, Vol.72, No. 5: pp. 1070 - 1087.

Hausman, Jerry. "Taxes and Labor Supply." In Handbook of Public Economics, Vol. 1,
Alan Auerbach and Martin Feldstein (eds.). Amsterdam: North Holland, 1985,
pp. 213 - 263.

Haveman, Robert H. Poverty Policy and Poverty Research: The Great Society and the
Social Sciences. Madison, WI: University of Wisconsin Press, 1987.

Hill, Carolyn J ., V. Joseph Hotz, Charles H. Mullin, and John Karl Scholz. “EITC
Eligibility, Participation and Compliance Rates for AFDC Households: Evidence
ﬁom the California Caseload.” Sacremento, CA: State of California, April 1999.

Holtzblatt, Janet, Janet McCubbin, and Robert Gillette. "Promoting Work through the
EITC." National Tax Journal, Vol. XLVII, No. 3, (Sept. 1994): pp. 591 - 607.

Hotz, V..Joseph and John Karl Scholz. "The Earned Income Credit.” Paper for the
NBER Conference on Means-Tested Transfers, July 15, 2000.

Internal Revenue Service. Statistics of Income —- 1998, Individual Income Tax Returns.
Washington, DC: 2001.

Johnson, Nicholas. “A Hand Up: How State Earned Income Tax Credits Help Working

Families Escape Poverty, 1999 Edition.” Washington, DC: Center on Budget and
Policy Priorities, November 1999.

206

Johnson, Nicholas and Ed Lazere. “A Rising Number of States Offer Earned Income Tax
Credits.” Washington, DC: Center on Budget and Policy Priorities, September
1998.

Killingsworth, Mark R. Labor Supply. Cambridge, England: Cambridge University
Press, 1983.

Lazear, Edward. “The Narrowing of Black-White Wage Differential is Illusory.”
American Economic Review, Vol. 69, (Sept. 1979): pp. 553 - 564.

Leighton, Linda, and Jacob Mincer. ”The Effects of Minimum Wages on Human Capital
Formation.” In The Economics of Legal Minimum Wages, edited by Simon
Rottenberg. Washington, DC: American Enterprise Institute for Public Policy
Research, 1981, pp. 88 - 123.

Mattila, J. Peter. “The Impact of Minimum Wages on Teenage Schooling and on the
Part-Time/Full-Time Employment of Youths.” The Economics of Legal Minimum
Wages, edited by Simon Rottenberg. Washington, DC: American Enterprise
Institute for Public Policy Research, 1981, pp. 61 - 87.

Mattila, J. Peter. “ Youth Labor Markets, Enrollments, and Minimum Wages.” Industrial
Relations Research Association Proceedings, 31, (1978): pp. 134 - 140.

Meyer, Bruce D. and Dan T. Rosenbaum. “Welfare, the Earned Income Tax Credit, and
the Labor Supply of Single Mothers.” The Quarterly Journal of Economics, Vol.
116, No. 3, (Aug. 2001): pp. 1063 - 1114.

Meyer, Bruce D. and Dan T. Rosenbaum. “Making Single Mothers Work: Recent Tax
and Welfare Policy and Its Effects.” National Tax Journal, Vol. LIII, No. 4, Part
2, (Dec. 2000): pp. 1027 - 1061.

Mincer, Jacob. Unemployment Effects of Minimum Wages. Journal of Political
Economy, Vol. 84, No. 4, Part2, 1976: pp. 887 - S104.

Mincer, Jacob. Schooling, Experience, and Earnings. National Bureau of Economic
Research and Columbia University Press, New York, NY: 1974.

Mofﬁtt, Robert. "Incentive Effects of the U.S. Welfare System: A Review." Journal of
Economic Literature, Vol. XXX, (March 1992): pp. 1 - 61.

Mroz, Thomas. "The Sensitivity of an Empirical Model of Married Women's Hours of
Work to Economic and Statistical Assumptions." Econometrica, Vol. 55, (July
1987): pp. 765 - 800.

207

Neumark, David, and William Wascher. “Do Minimum Wages Fight Poverty?”
Economic Inquiry, Vol. 40, No. 3, (2002): pp. 315 - 333.

Neumark, David, and William Wascher. “Minimum Wages and Training Revisited.”
Journal of Labor Economics, Vol. 19, No. 3, (2001): pp. 563 - 595.

Neumark, David and William Wascher. “Using the EITC to Help Poor Families: New
Evidence and a Comparison with the Minimum Wage.” National Tax Journal,
Vol. LIV, No. 2, (June 2001): pp. 281 - 317.

Neumark, David, Mark Schweitzer, and William Wascher. “The Effects of Minimum
Wages on the Distribution of Family Incomes: A Non-Parametric Analysis.”
Working Paper No. 6536. Cambridge, MA: NBER, updated June 2000.

Neumark, David, and William Wascher. “Minimum Wage Effects and Low-Wage Labor
Markets: A Disequilibrium Approach.” Working Paper, Michigan State
University, March 1997.

Neumark, David, and William Wascher. “The Effects of Minimum Wages on Teenage
Employment and Enrollment: Evidence from Matched CPS Surveys.” Research
in Labor Economics, Vol. 15, 1996: pp. 25 - 63.

Romich, Jennifer L., and Thomas Weisner. “How Families View and Use the EITC:
Advance Payment versus Lump Sum Delivery.” National Tax Journal, Vol. 7,
No. 3, (Dec. 2000): pp. 1245 - 1265.

Rosen, Sherwin. “Learning and Experience in the Labor Market.” Journal of Human
Resources, Vol. VH, No. 3, (1972): 326 - 342.

Schoeni, Robert F., and Rebecca Blank. “The Effects of Welfare Reform and Welfare
Waivers on Employment, Income, Poverty, and Family Structure.” Working
Paper, Berkeley, CA: February 2000.

Scholz, John Karl. "The Earned Income Tax Credit: Participation, Compliance, and
Antipoverty Effectiveness." National Tax Journal, Vol. XLVH, No. 1, (March
1994): pp. 63 - 87.

Smeeding, Timothy M., Katherin Ross Phillips, and Michael O’Connor. “The EITC:
Expectation, Knowledge, Use, and Economic and Social Mobility.” National Tax

Journal, Vol. LIII, No. 4, Part 2, (Dec. 2000): pp. 1187 - 1209.

Triest, Robert K. "The Effect of Income Taxation on Labor Supply in the United States."
Journal of Human Resources, Vol. XXV, No. 3, (Summer 1990): pp. 491 - 516.

U.S. Bureau of the Census. CURRENT POPULATION SURVEYS, MARCH 1962-

208

U.S.

U.S.

U.S.

U.S.

U.S.

U.S.

U.S.

U.S.

U.S.

U.S.

1997 [machine-readable data ﬁles]/conducted by the Bureau of the Census for the
Bureau of Labor Statistics. Washington: Bureau of the Census [producer and
distributor], 1962-1997. Santa Monica, CA: Unicon Research Corporation
[producer and distributor of CPS Utilities], 1998.

Committee on Ways and Means. Green Book. Washington, DC: U.S. Government
Printing Ofﬁce, various years.

Department of Commerce. Bureau of the Census. Historical Poverty Tables.
Washington, DC: Census Bureau Web site
www.census.gov/hhes/povertv/histpov/hstpov21 .html.

Department of Commerce. Bureau of Economic Analysis. Survey of Current
Business. Washington, DC: U.S. Government Printing Ofﬁce, various years.

Department of Commerce. Bureau of Economic Analysis. State Personal Income,
1929-8 7: Estimates and a Statement of Sources and Methods. Washington, DC:
U.S. Government Printing Ofﬁce, July 1989.

Department of Health, Education, and Welfare. Social Security Administration. Aid
to Families with Dependent Children. Washington, DC: U.S. Government
Printing Ofﬁce, March 1978.

Department of Health and Human Services. Administration for Children and
Families. 2000 TANF Report to Congress. Washington, DC: U.S. Government
Printing Ofﬁce, August 2000.

Department of Labor. Minimum Wage and Overtime Hours Under the Fair Labor
Standards Act. Washington, DC: Department of Labor, January 2001.

Department of Labor. Monthly Labor Review. Washington, DC: Department of
Labor, 1976 - 1998.

Department of Labor. Minimum Wage and Maximum Hours Standards Under the
Fair Labor Standards Act. Washington, DC: Department of Labor, 1983.

Ofﬁce of Management and Budget. FY 2002 Economic Outlook, Highlights from
FY 1994 to FY2001, FY 2002 Baseline Projections. Washington, DC: U.S.
Government Printing Ofﬁce, January 2001.

Ventry, Dennis J. "The Collision of Tax and Welfare Politics: The Political History of

the Earned Income Tax Credit, 1969-99." National Tax Journal, Vol. LIII, No.4,
Part 2, (Dec. 2000): pp. 983 - 1026.

Welch, Finis. “Minimum Wage Legislation in the United States.” Economic Inquiry,

209

Vol. 12, No. 3, (1974): pp. 285 - 318.

Wooldridge, Jeffrey M. Economietric Analysis of Cross Section and Panel Data. The
MIT Press, Cambridge, MA: 2002.

210

 

 

 

      
 
 

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

1111111111111111111111 1111

. 31293 026