Making the Work-Based Safety Net Work Better: Forward-Looking Policies to Help Low-Income Families 0871544660, 9780871544667

Citation preview

Making the Work-Based Safety Net Work Better

Making the Work-Based Safety Net Work Better Forward-Looking Policies to Help Low-Income Families

Carolyn J. Heinrich and John Karl Scholz, Editors

Russell Sage Foundation  New York

The Russell Sage Foundation The Russell Sage Foundation, one of the oldest of America’s general purpose foundations, was established in 1907 by Mrs. Margaret Olivia Sage for “the improvement of social and living conditions in the United States.” The Foundation seeks to fulfill this mandate by fostering the development and dissemination of knowledge about the country’s political, social, and economic problems. While the Foundation endeavors to assure the accuracy and objectivity of each book it publishes, the conclusions and interpretations in Russell Sage Foundation publications are those of the authors and not of the Foundation, its Trustees, or its staff. Publication by Russell Sage, therefore, does not imply Foundation endorsement. BOARD OF TRUSTEES Mary C. Waters, Chair Kenneth D. Brody W. Bowman Cutter, III Christopher Edley Jr. John A. Ferejohn Larry V. Hedges

Kathleen Hall Jamieson Melvin J. Konner Alan B. Krueger Cora B. Marrett Nancy Rosenblum

Shelley E. Taylor Richard H. Thaler Eric Wanner

Library of Congress Cataloging-in-Publication Data Making the work-based safety net work better : forward-looking policies to help lowincome families / Carolyn J. Heinrich and John Karl Scholz, editors. p. cm. Includes bibliographical references and index. ISBN 978-0-87154-466-7 (alk. paper) 1. Working poor—Government policy—United States. 2. Poor families—Government policy—United States. I. Heinrich, Carolyn J. II. Scholz, John Karl. HD8072.5.M34 2009 362.5’840973—dc22 2008050601 Copyright © 2009 by Russell Sage Foundation. All rights reserved. Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Reproduction by the United States Government in whole or in part is permitted for any purpose. The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences—Permanence of Paper for Printed Library Materials. ANSI Z39.48-1992.

RUSSELL SAGE FOUNDATION 112 East 64th Street, New York, New York 10065 10 9 8 7 6 5 4 3 2 1

We dedicate this book to our spouses, Kurt Heinrich and Melissa Auchard Scholz, and to our children, Victoria, Nathan, and Lauren Heinrich, and Libby, Kate, and Karly Scholz.

Contents

Chapter 1

ABOUT THE AUTHORS

ix

ACKNOWLEDGMENTS

xi

MAKING THE WORK-BASED SAFETY NET WORK BETTER

Carolyn J. Heinrich and John Karl Scholz PART I

Chapter 2

CHALLENGES FACED BY ADULTS IN ACHIEVING SELF-SUFFICIENCY THROUGH WORK

PART II

Chapter 4

PART III

Chapter 6

25

ON WORK AND HEALTH AMONG THE AMERICAN POOR

Jayanta Bhattacharya and Peter Richmond

79

IMPROVING CHILDREN’S CHANCES OF BECOMING SELF-SUFFICIENT ADULTS

115

PARENTAL PATHWAYS TO SELF-SUFFICIENCY AND THE WELL-BEING OF YOUNGER CHILDREN

Greg J. Duncan, Lisa Gennetian, and Pamela Morris Chapter 5

23

ALTERNATIVE LABOR-MARKET POLICIES TO INCREASE ECONOMIC SELF-SUFFICIENCY: MANDATING HIGHER WAGES, SUBSIDIZING EMPLOYMENT, AND INCREASING PRODUCTIVITY

David Neumark Chapter 3

1

117

SCHOOL REFORMS AND IMPROVED LIFE OUTCOMES FOR DISADVANTAGED CHILDREN

David N. Figlio

149

ADDRESSING BARRIERS TO SELF-SUFFICIENCY FOR PARTICULARLY DISADVANTAGED GROUPS

183

THE IMPACT OF INCARCERATION ON THE EMPLOYMENT OUTCOMES OF FORMER INMATES: POLICY OPTIONS FOR FOSTERING SELF-SUFFICIENCY

Steven Raphael

185

Making the Work-Based Safety Net Work Better

Chapter 7

PART IV

Chapter 8

THE GROWING PROBLEM OF DISCONNECTED SINGLE MOTHERS

Rebecca M. Blank and Brian K. Kovak

227

POLICY IDEAS FROM OTHER COUNTRIES AND THE POLITICS OF CHANGING THE WORK-BASED SAFETY NET

259

BEYOND THE SAFETY NET: SUPPORTING THE ECONOMIC SECURITY OF WORKING-POOR FAMILIES

Marcia K. Meyers and Janet C. Gornick Chapter 9

viii

/

261

THE POLITICS OF LOW-INCOME FAMILIES IN THE UNITED STATES

R. Kent Weaver

292

INDEX

339

About the Authors

CAROLYN J. HEINRICH is director of the La Follette School of Public Affairs, professor of public affairs and affiliated professor of economics, and associate director of research and training at the Institute for Research on Poverty at the University of Wisconsin, Madison. JOHN KARL SCHOLZ is professor of economics at the University of WisconsinMadison and research associate at the National Bureau of Economic Research.

JAYANTA BHATTACHARYA is associate professor of medicine at Stanford University and faculty research fellow at the National Bureau of Economic Research. REBECCA M. BLANK is Robert S. Kerr Senior Fellow at the Brookings Institution and research associate with the National Bureau of Economic Research. GREG J. DUNCAN is Distinguished Professor of Education at the University of California, Irvine. DAVID N. FIGLIO is Orrington Lunt Professor of Education, Social Policy and Economics and faculty fellow at the Institute for Policy Research at Northwestern University, and research associate at the National Bureau of Economic Research. LISA GENNETIAN is senior research director of economic studies at the Brookings Institution. JANET C. GORNICK is professor of political science and sociology at the Graduate Center, City University of New York and director of the Luxembourg Income Study. BRIAN K. KOVAK is a Ph.D. candidate in the Department of Economics at the University of Michigan. MARCIA K. MEYERS is professor of social work and public affairs at the University of Washington and director of the West Coast Poverty Center.

About the Authors

PAMELA MORRIS is codirector of the policy area on family well-being and children’s development at MDRC. DAVID NEUMARK is professor of economics at the University of California, Irvine, senior fellow at the Public Policy Institute of California, research associate at the National Bureau of Economics Research, and research fellow at IZA. STEVEN RAPHAEL is professor of public policy at the Goldman School of Public Policy at the University of California, Berkeley. PETER RICHMOND is research assistant at the Center for Health Policy and Center for Primary Care and Outcomes Research (CHP/PCOR) at Stanford University. R. KENT WEAVER is professor at the Public Policy Institute at Georgetown University and senior fellow in governance studies at the Brookings Institution.

x

/

Acknowledgments

The chapters in this volume were commissioned for a conference, “Pathways to Self-Sufficiency,” held at the Institute for Research on Poverty at the University of Wisconsin, Madison, September 6 and 7, 2007. We are extremely grateful to the authors for their insightful papers. We also thank the discussants at that conference—Harry Holzer, John Cawley, Bobbi Wolfe, Diane Whitmore Schanzenbach, Michael Stoll, Jim Ziliak, Tim Smeeding, and Tom Gais—and other conference participants, whose comments and insights helped to shape this book. We are very grateful to the Russell Sage Foundation, the Annie E. Casey Foundation, the Smith Richardson Foundation, the Office of the Assistant Secretary for Planning and Evaluation at the U.S. Department of Health and Human Services, and the University of Wisconsin’s Institute for Research on Poverty for financially supporting this undertaking. We are also thankful for the comments of Ron Haskins, the collaboration of Tom Kaplan, and the encouragement of Maria Cancian, the director of the Institute for Research on Poverty, during the planning and execution of this project. Finally, this endeavor would not have been possible without the assistance and professionalism of the institute staff, particularly Emma Casper, Dawn Duren, Deborah Johnson, Robin Snell, Bill Wambach, and Coreen Williams, who remain the best.

Chapter 1 Making the Work-Based Safety Net Work Better Carolyn J. Heinrich and John Karl Scholz

T

he idea that work is the gateway to self-sufficiency and the way out of poverty is a simple, powerful message. Work-based welfare reform aligns the structure of the safety net with the central values of Americans who are not on welfare. As R. Kent Weaver points out in his political analysis in this volume (see chapter 9), a key element of continuity in American attitudes concerning policy toward low-income families is the emphasis on and expectation that the able-bodied should work. Indeed, it is difficult to see how the safety net for disadvantaged families could fail to embrace work when over 75 percent of women and 91 percent of men between the ages twenty-five and fifty-four are in the workforce.1 Work-based policy reform not only has wide-ranging political appeal, but also is based on a concept that is supported by academic research. One of the earliest and best-known studies of a “work-first” program, the randomized-controlled trial of California’s Greater Avenues to Independence (GAIN) program, showed that participants in the Riverside program, which strongly emphasized immediate work, achieved considerably larger short-run gains in earnings than did those in other GAIN sites that invested more in individual skill development. In addition, the Riverside GAIN program had the lowest costs of the six GAIN sites (MDRC 1994). These findings reached the press and academic forums just as public debate of welfare reform proposals was climaxing, so it was perhaps not surprising that ending “the dependence of needy parents on government benefits” through job preparation and work was a fundamental goal of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996.2 As appealing as the rhetoric of self-sufficiency is, it collides with an exceptionally difficult reality for individuals and families who struggle to escape poverty even as they work, or who cycle in and out of work and welfare. To satisfactorily

/

1

Making the Work-Based Safety Net Work Better

achieve the ultimate goal of PRWORA, heads of households have to find and hold jobs that allow them to nurture the physical and psychological well-being of children and other adults in the household. At the same time, it appears that the ability of these adults to support their families at anything close to poverty-line earnings is limited.3 The labor-market skills of a substantial fraction of adult welfare recipients are notably modest. Just prior to welfare reform, 42 percent of adults in families that received welfare benefits at some point during the year had less than a high school diploma. The overall employment rate of prime-age workers with less than a high school diploma was only 36 percent. Not surprisingly, as the reforms of the 1990s unfolded, many were skeptical that welfare recipients could get and hold jobs, barring major investments in child care, health insurance, transportation assistance, and training.4 From this vantage point, the subsequent reduction in welfare caseloads, from a peak of 5.1 million families in March 1994 to 2.1 million families in March 2001 (and holding steady through March 2004), was nothing short of extraordinary. Some of the sharp reduction in welfare caseloads was undoubtedly the result of purposeful efforts to discourage eligible families from receiving benefits (“diversion” efforts), but few dispute that there have also been far-reaching changes in the way states implement and manage programs, most notably the increased focus on work (Grogger and Karoly 2005; Mead 2004). In addition, until 2001 the economy was very strong, generating more than 17 million new jobs and pushing unemployment rates to their lowest levels since 1969. Wages grew throughout the income distribution, as they did not during the economic expansions in the 1980s, while real Earned Income Tax Credit (EITC) benefits more than doubled. Poverty among single-parent (mostly female-headed) households also declined precipitously until the recession of 2001. In effect, this coincidental timing of welfare changes, the strong economy, and EITC expansions paved the way for many more low-skilled people to enter the workforce and improve their economic well-being, although the relative contributions of these factors to their increased employment and reduced poverty is still a matter of debate.5 Through PRWORA and its 2006 reauthorization (under the Deficit Reduction Act of 2005), U.S. policymakers are clearly committing to a work-based safety net. These developments in policy and low-wage labor markets over the past fifteen years motivate the focal question posed in this volume: What further changes are needed to meet the goals of increasing self-sufficiency and reducing poverty among low-income and disadvantaged families and individuals? Lawrence M. Mead (2007) argues that the welfare reform model with stringent work requirements has been so successful in getting low-income mothers to work that it should be adapted for low-income males as well. In other words, it would compel them to internalize the obligation to work.6 Others, however, express concern that a work-based safety net will continue to push low-skilled individuals into deadend jobs that offer little prospect of providing a permanent pathway out of poverty for them and their families. In fact, a recent reanalysis of the GAIN data over a longer (nine-year) follow-up period showed that the short-term advan-

2

/

Making the Work-Based Safety Net Work Better

tages of a work-first strategy dissipated over time, and that in the long run, programs stressing skills development yielded higher employment rates and significantly better economic outcomes for participants (Hotz et al. 2006).

IMPETUS FOR CHANGES IN SAFETY NET POLICY Recent research, including the work presented in this volume, suggests there are fragilities, if not glaring holes, in the work-based safety net for which innovative policy changes that could improve the well-being of low-income families must be considered. Among the more worrisome facts, a substantial body of prior research has examined average earnings and growth in earnings among families leaving welfare after the reforms and concluded that despite the exceptionally strong economy and rapid job growth of the later 1990s, the earnings of welfare leavers were still very low and job turnover was high. In their study of welfare reform experiences in six large cities during the 1996-to-1999 economic expansion, Christopher T. King and Peter R. Mueser (2005) observed an increase in the number of individuals who left welfare prior to finding employment. They suggest that although one might argue that most of these individuals were drawn off of welfare by new employment opportunities, it is more likely that their exit from welfare reflected other factors, including the more stringent conditions for welfare receipt. They concluded that welfare reform did not change the reality that labormarket demand for this group “seldom provides earnings sufficient to pull a family out of poverty,” and that individuals were probably not made better off by reform (King and Mueser 2005, 165). It has also been argued, however, that to more fully understand the impact of welfare reform, one should consider the welfare “averters,” that is, those who do not enter the welfare rolls following welfare reform. Jeffrey Grogger and Lynn A. Karoly (2005) reported that, at least in the period before time limits became binding, observational studies show that the income of disadvantaged women rose and their poverty rate fell. They suggest that one possible explanation for these gains is that welfare reform contributed to increases in the income of those who were deterred from entering welfare by the reforms, and that these individuals would not be captured in the experimental studies or leaver studies focusing on those applying for or receiving welfare. They also acknowledge, however, that the observational studies that find rising incomes and falling poverty rates among these individuals may simply be overstating the effects of welfare reform due to inadequate controls for intervening factors, such as other changes in the economy or the effects of EITC expansions, minimum wage increases, or other policy changes. Wisconsin is a particularly good place to examine additional evidence about the antipoverty implications of employment for families leaving welfare, given its exceptionally strong economy in the late 1990s and its early reform efforts, which placed great emphasis on work but also provided strong work supports. Researchers affiliated with the Institute for Research on Poverty (IRP; based at the

/

3

Making the Work-Based Safety Net Work Better

University of Wisconsin, Madison) studied the earnings of two groups of women leaving welfare in Wisconsin (Cancian et al. 2000). Among those who left welfare in the fourth quarter of 1995, average earnings for the year after they left were $9,108 (the median was $8,608). And although average earnings rose to $10,294 in the second year after exit and to $11,450 in the third year (all in 1998 dollars), most families still had average earnings significantly below the poverty line three years after leaving welfare.7 Even incorporating earnings supplements from the federal and Wisconsin EITCs, only 37.4 percent of families that had left welfare in the fourth quarter of 1995 in Wisconsin had after-tax (and after-EITC) earned incomes exceeding the poverty line in 1998.8 Data from California showed comparably slow earnings growth for families trying to make the transition from welfare to work (Hotz, Mullin, and Scholz 2002). Still another study of the rate of wage growth for less-well-educated women relative to other groups cautioned that “work experience is not a magic bullet. . . . Evidence indicates that low-skilled workers will not have huge wage gains from work experience” (Gladden and Taber 2000).9 And, perhaps most disturbing, Rebecca Blank and Brian Kovak show in “The Growing Problem of Disconnected Single Mothers” (chapter 7, this volume) that although overall poverty in single-mother families changed little from 2000 to 2005, extreme poverty rose by 2.7 percentage points between these years, suggesting that there were more extremely poor female-headed families in 2005 than there were in 2000. Briefly summarizing, employment is an important first step for families trying to achieve self-sufficiency, but the earnings of most individuals who left welfare during the strong economy of the 1990s were still well below the poverty line even many years later.10 And although a majority of studies have focused exclusively on the earnings of individuals rather than the total income of families, the limited set of more recent studies that have attempted to account for the earnings and benefits available to families suggest that the conclusions we draw from the experiences of individuals hold more broadly (Meyer and Cancian 1998; Primus et al. 1999). Hence, the degree to which work will be the primary antidote to poverty will depend greatly on the ability of low-skilled people to maintain employment that over time offers a progression of income that allows families to be self-sufficient. To date, the evidence in support of a strictly work-focused approach is hardly encouraging.

THE GOALS OF THIS VOLUME The chapters in this volume focus on self-sufficiency, a concept that requires further discussion in light of the various meanings ascribed to it by different people. For some scholars, it may suggest a situation where families and individuals get by solely from the incomes earned in the labor market and goods produced at home. Gueron (1990, 80) noted that in the long history of the concept of selfreliance, policymakers and scholars have tended to divide the poor into two groups: “those who are able and expected to work and those who are not,” with the

4

/

Making the Work-Based Safety Net Work Better

idea that assistance should be provided only to the latter (for example, the aged and severely disabled). We do not concur with this narrow conception that those who are able to work should be fully independent of government assistance, as it largely ignores the pervasive role that government plays in the lives of all citizens. Middle- and upper-income families benefit from substantial housing and other tax-based subsidies. Lower-income households benefit from government-provided work supports such as child care subsidies and the Earned Income Tax Credit (EITC). An alternative and considerably broader conception of self-sufficiency, articulated primarily by European scholars and policymakers, is based on the idea of entitlement to a “customary minimum standard of living” over the life course. As Gardiner (2000, 680) elaborates, self-sufficiency requires not only employment with adequate pay but also time for family members to provide for “family care needs” that cannot be met by public or private providers; an “adequate level of social security allowances for children, unemployment, maternity, parental leave, sickness, disability and retirement”; and affordable services to substitute for domestic labor and family care. This perspective suggests that the government has an obligation to secure each of these “elements of self-sufficiency” for its citizens in order to provide for them a reasonable standard of living. We do not ascribe to this viewpoint either, because we do not see it as politically or economically viable to define the United States government’s obligation as being this extensive. Instead, we adopt a more flexible, intermediate conception of self-sufficiency, which we define as families’ having an above-poverty standard of living through their labor-market earnings and associated public benefits (whether the EITC, food stamps, S-CHIP, housing, or something else) for which they are eligible. In advancing this working definition we make two further observations. First, there is an inherent and long-recognized tension between unqualified self-sufficiency— the absence of any public assistance—and poverty reduction (Ellwood 1988; Garfinkel and McLanahan 1986; Gueron 1990). Second, current measures of the poverty line are very outdated, as discussed by Constance Citro and Robert Michael (1995). According to Citro and Michael, the “above-poverty” standard of living referred to would appropriately account for public benefits. We concur with this assessment. In assembling this volume, the editors and contributors perceived a dearth in the academic and policy literatures of forward-looking discussions of the workbased safety net. Much academic writing, particularly research appearing in scholarly journals, tends to be “backward looking,” in the sense that it uses past data to interpret the effects of different policies or to test behavioral theories. In addition, we find little scholarly writing specifically discussing how the safety net could evolve in ways that would enhance the ability of low-skilled households to be self-sufficient, and scant systematic policy experimentation with approaches to promoting self-sufficiency.11 Of course, an important challenge in looking forward is to not only identify policies that help people get and retain jobs as well as increase earnings but also to consider their effects on children or other family members.

/

5

Making the Work-Based Safety Net Work Better

The authors in this volume embrace the challenge of looking forward, examining different safety-net domains that are fragile, inadequate, or in of need revamping. They describe how a set of policies and institutions could evolve to enhance the self-sufficiency of poor families. Asking scholars to write forward-looking chapters runs the risk of conjuring grandiose ideas that might not have the remotest chance of entering policy debates and political discussions. Recognizing this, we strove to arrive at a set of well-grounded ideas and policy prescriptions arising from the work of the outstanding scholars who contributed to this volume. Owing largely to the high levels of engagement each of these scholars has with current policy debates and the major research contributions they have made to these subject areas in the past, this volume presents a provocative blend of both incremental ideas that could reasonably enter political and policy discussions immediately and bolder ideas that may sit until consideration of more extensive or systemic changes in the safety net are politically and economically feasible.

A BRIEF OVERVIEW OF KEY FINDINGS The book is divided into four parts. Part I is “Challenges Faced by Adults in Achieving Self-Sufficiency Through Work.” A work-based safety net must facilitate access to stable employment and adequate earnings for people to achieve self-sufficiency. At the same time, work may have important (positive or negative) implications for individual health, as may the lack of or loss of employment. The authors of the chapters in part I examine policies that foster employment and the effects that low-wage employment may have on health. In chapter 2, David Neumark reviews the evidence on labor-market policies— particularly the minimum wage, EITC, wage subsidies, and school-to-work transition programs—that are aimed at promoting work among low-income individuals and families and at increasing their ability to achieve an adequate standard of living from their participation in the labor market. He considers three basic policy strategies for increasing the earnings of low-income families: directly mandating higher wages through minimum or living wages; promoting employment through demand-side or supply-side incentives for work (such as the EITC and wage subsidies), and increasing human capital through education and training. Following an extensive review of the literature and evidence and careful consideration of the complex effects of these policies, he concludes that the minimum wage is an ineffective policy to promote economic self-sufficiency, as it is likely to reduce employment, and the benefits are poorly targeted.12 He gives the earned income tax credit a more positive evaluation, writing: “There seems fairly compelling evidence that a more generous EITC boosts employment of single mothers and in so doing raises incomes and earnings of low-income families.” Among policies that have seen less attention, Neumark suggests that wage subsidies targeted to individuals, rather than their employers, are worth further consideration, though he raises concerns about administrative issues that arise in implementing

6

/

Making the Work-Based Safety Net Work Better

these policies. These policies could take the form of subsidies to earnings or expanding the EITC to childless workers. In considering policies for raising human capital, Neumark focuses on the new but growing literature on school-to-work programs that generates some evidence of beneficial effects of these programs—in particular, an increase in skill formation and college attendance among youth. At the same time, one needs to be careful to balance school-to-work initiatives with the need for academic preparation. School-to-work programs also do not address the concerns of skill deficits among workers beyond school age. Ongoing training efforts need to encourage innovative programs, embrace rigorous program evaluation, discard unsuccessful programs, promote successful ones, and be mindful of scale, in the sense that effective programs must have the potential to be disseminated effectively to other jurisdictions. In chapter 3, Jayanta Bhattacharya and Peter Richmond examine the consequences of work-based safety-net policies on individual health and well-being. The effectiveness of a work-based safety net is based on the expectation that individuals will be capable of participating in the labor market over the majority of their adult lives. To the extent that work done by low-skilled, low-income workers harms their physical or psychological health, it may be difficult for them to work consistently and maintain self-sufficiency. At the same time, unemployment can likewise have damaging effects on physical and mental health. Thus, the effects of working on health are a critical but complex and poorly understood topic in welfare reform. Chapter 3 provides an introduction to and analysis of this understudied issue, in addition to suggestions for policy reforms to promote the health of poor workers. If work damages health for workers in low-skilled jobs, then the prospects for a work-based safety net to increase self-sufficiency would seem to be much less promising. And because the relationship between work and health could run in either direction, efforts to disentangle causal effects are complicated. Work can be stressful and some jobs are dangerous. In addition, work may consume time that is needed to nurture physical health and to maintain relationships with friends and family, which could also be a factor affecting psychological health. At the same time, not having a job can also be stressful and psychologically debilitating. Yet prior research suggests that workers’ health status actually improves during economic downturns. This issue clearly strikes to the heart of the soundness of a work-based safety net, but unlike other chapters in this volume, there is not an extensive, prior literature to be presented that examines these issues. Bhattacharya and Richmond find that poor workers tend to use leisure more effectively than nonworkers in addressing their physical and psychological health, but work reduces the amount of leisure people have, thus jeopardizing health-promoting activities. Policy initiatives discussed in the chapter include those that may increase the time for activities that promote physical and psychological health, including approaches that reduce commuting time, improve child-care options, or allow work schedules to be more flexible. A noteworthy aspect of the chapter is its discussion of serum mark-

/

7

Making the Work-Based Safety Net Work Better

ers of chronic disease, which we hope will increase recognition of biomedical measures of well-being that could be profitably used in other poverty-related research efforts.13 Part II, “Improving Children’s Chances of Becoming Self-Sufficient Adults,” examines the issues raised by a work-based safety net for young and school-age children. Low-skilled individuals, not surprisingly, are rarely working to support just themselves; they typically have family and other obligations. Demands of the labor market may interfere with responsibilities for the care and well-being of children as well as their performance in school. In chapter 4, Greg Duncan, Lisa Gennetian, and Pamela Morris explore the design features of work-based antipoverty programs that appear to have the most beneficial (or least detrimental) effects on young children. Transitions from parental welfare receipt to work may benefit children if the children receive highquality child care, and if parents become better role models, develop a better sense of maternal and paternal self-esteem and sense of control, and perhaps, foster career advancement and greater household consumption. Of course, greater induced employment may instead overwhelm already stressed parents, force children into less desirable child-care arrangements, decrease parental supervision of children, and for those unable to comply with program rules, deepen family poverty. Duncan, Gennetian, and Morris, using variation created by a series of well-implemented randomized social experiments, examine the causal pathways through which work-based antipoverty policies may affect child well-being. They find considerable evidence that policies that increase family income (earnings plus benefits) and that lead to greater use of center-based child care can improve child achievement. There is very little evidence that changes to child well-being occur through policies affecting maternal mental health or parenting practices.14 Chapter 4 nicely highlights a tension in the design of welfare programs as they affect children. Such programs can increase parental self-sufficiency (while retaining a work-first focus), provide few benefits to children, and save costs. Or, at greater cost, they can use earnings supplements to increase parental employment, raise family income, and provide benefits to children. Scholars and policy analysts have long been concerned about intergenerational aspects of poverty. Indeed, one of the rationales for the work requirements of the 1996 welfare reform was that by requiring parents to be in the labor force they would provide positive role models for their children, thus reducing the likelihood that their children would be poor. To take seriously concerns about intergenerational linkages of poverty, however, it is imperative also to be concerned with children’s educational attainment and the role of public schools in educating children from poor families. In chapter 5, David Figlio provides a forward-looking analysis of developments in K-through-twelve education reform and how they might help children from low-income families to overcome the disadvantages with which they typically begin school. Figlio describes the strong, growing relationship between educational attainment and self-sufficiency in the United States. For example, in 2000, over 25 percent of all adults with less than high school education but only 2.5 percent of col-

8

/

Making the Work-Based Safety Net Work Better

lege graduates had incomes below the poverty level. The evidence also shows that children from low-income families begin school at a considerable disadvantage and end school with lower educational attainment. Despite the many forms of remediation that have been used to reduce the gap between advantaged and disadvantaged children, these gaps grow as children progress through school, resulting in significantly lower human capital acquisition that depresses the likelihood of children from low-income families of achieving self-sufficiency. Figlio notes that one important reason we have made so little progress in closing education outcome gaps is that schools provide just some of the many inputs into children’s learning and success in school. Figlio contends that although we should demand that schools provide equality of educational opportunities for all children, it is unreasonable to expect equality of educational outcomes. He reviews the many approaches to improving educational quality and outcomes that have been proposed or tested—including schoolfinance reforms, reduced class sizes and student-teacher ratios, teacher compensation and strategies for improving teacher quality, and school choice—and argues that the key to improving educational outcomes for disadvantaged children is to improve teacher quality. In light of the substantial differences in teacher quality both within and between schools, Figlio sees considerable potential for using incentives to encourage high-quality teachers to move to and remain in schools serving large numbers of disadvantaged students. At the same time, he cautions that the specific incentive policies need to be carefully considered and implemented, given the difficulties in measuring teacher quality and the potential negative effects of accountability mechanisms (for example, increased teaching to the test) that aim to reward teachers for student performance. Part III, “Addressing Barriers to Self-Sufficiency for Particularly Disadvantaged Groups,” examines criminal-justice policy and the problems that arise with the hardest to employ. Youths and adults in poor families are at a disproportionately high risk of interaction with the criminal-justice system and subsequent incarceration. Moreover, no matter how well designed labor-market policies are, some individuals are unlikely to be able to achieve economic self-sufficiency. In 2001, the lifetime probability of a male spending time in jail or prison was 5.9 percent for whites, 32.2 percent for blacks, and 17.2 percent for Hispanics. One of the particularly troubling weaknesses of the work-based safety net is its inadequate provision for dealing with the increasingly difficult challenge of accommodating the large number of men and women who are or were in prisons and jails in the United States. Extensive social science evidence documents the employment difficulties associated with prolonged exposure to the criminal justice system. These issues are discussed in Chapter 6 by Steven Raphael, who offers constructive, tractable suggestions for policy changes that could ease the transition of men and women back into non-institutionalized society and the labor market. It is hard to imagine healthy communities in which up to one-third of poorly educated working-age African American men may be in jail or prison on a given day and a considerably larger fraction are or have previously been under the jurisdiction of the criminal justice system. Most of these men have child-support

/

9

Making the Work-Based Safety Net Work Better

orders and little income that they can use to meet their financial responsibilities to their children. They face extremely poor employment outcomes, yet there are few existing federal or state programs that are designed to foster economic self-sufficiency of single, noncustodial parenting men, particularly those with a criminal record. There are several policy changes that could limit the adverse consequences of corrections policies on poor, minority communities. These would include eliminating federal bans on the participation of certain convicted felons in various public assistance programs; rationalization of federal, state, and local employment bans to allow for greater consideration of the details of particular cases; legislative guidance on how employers may or may not consider the criminal history of job applicants; and, for state programs, expunging criminal records of former inmates who exhibit sustained desistance from criminal activity and meet other benchmarks of responsible postrelease behavior. Raphael also argues for more proactive efforts to curb the inflow of youth into the system and to reduce the number of lives and families damaged by incarceration. In chapter 7, Rebecca Blank and Brian Kovak discuss the prevalence and problems of “disconnected” workers in the United States—those who report periods without earnings or public assistance benefits. As the safety net has evolved to emphasize work, employment and the earnings of low-income single mothers have increased. At the same time, Blank and Kovak observe, these developments have made assistance less available to those who are unsuccessful in securing or retaining employment and subsequently find themselves destitute. This is a longrecognized but inescapable tension that arises with a work-based antipoverty strategy: we wish to provide a humane level of benefits to low-income families, but in doing so we create incentives for persons to utilize government assistance rather than move into employment. Recognizing this problem and designing a system that adequately accommodates the needs of these individuals and families, without undermining incentives for work, is a central challenge for a workbased safety net. After documenting empirical facts about disconnected workers, Blank and Kovak discuss policy developments that could address the central design challenge for the work-based safety net. One possibility would be to expand noncash, means-tested programs and redouble efforts to enroll disconnected families and individuals in these programs. Another would be to expand Supplemental Security Income to allow for temporary or partial coverage. A third approach would be to create special programs that help highly disadvantaged single mothers who are having problems finding stable employment. The final policy option discussed in the paper is to remove barriers in TANF (Temporary Assistance for Needy Families) that make it difficult for states to provide ongoing support to disconnected mothers. U.S. policymaking does not occur in a vacuum and should be informed by the welfare reform efforts of other industrialized countries grappling with similar issues, although clearly the starting points for such reforms will differ. Moreover, all policy decisions are made in a specific political context, which affects policy

10

/

Making the Work-Based Safety Net Work Better

design, funding, and implementation. The authors of the chapters in part IV, “Policy Ideas from Other Countries and the Politics of Changing the Work-Based Safety Net,” consider these issues. In chapter 8, Marcia K. Meyers and Janet C. Gornick describe and draw evidence from recent social-welfare policy developments in several European countries that may point us toward policy alternatives for more effectively supporting the working poor. Given that employment loss and earnings instability are likely to be an enduring challenge for most poor families, Meyers and Gornick argue that current policies need to go beyond the conception of a safety net to develop a more comprehensive and accessible system of social and employment supports that assures the economic security of working but poor families over time. A more comprehensive, accessible set of social benefits and employment supports would reduce the number of disconnected workers or families. But as Meyers and Gornick document, there are more working families with children who are poor than there are families with children and no labor-market earnings.15 Looking to the countries of northern Europe for a model, they make the case that there is considerably more that the U.S. government could do to help families with low-earning parents achieve greater economic security. They focus on three types of policies with the potential to better support the efforts of families to attain an abovepoverty standard of living with their labor market earnings and public benefits: income benefits, paid family leave, and subsidized child care. Meyers and Gornick show that the United States performs considerably more poorly than its European peers in reducing child poverty rates through taxes and transfers and is particularly austere toward working-age families with children, who receive close to half of their supports as targeted, means-tested, or timelimited assistance. In addition, they note, there are no “institutional bridges” between means-tested welfare and other forms of assistance available to higherincome families. In other words, low-income families will lose their financial supports long before they earn enough to benefit from some of the subsidies available through tax provisions and employment, such as the home mortgage interest deduction and tax subsidies to employers for the provision of health insurance and pension arrangements. Recognizing the political and financial limits of policy reforms in this area, Meyers and Gornick argue for both bold and incremental changes to current policies, including a more inclusive unemployment insurance system, more generous, directly-funded child care assistance, and paid family leave financed through social insurance that distributes the burden more evenly between employees and employers. In general, the chapters in this volume take a traditional policy analysis approach to key issues critical to strengthening the work-based safety net and to identifying promising new policies for enhancing self-sufficiency. That is, they define the issue, survey the literature, analyze the problems, and offer informed suggestions for constructive approaches to addressing the problems. Of course, the likelihood of realizing beneficial, forward-looking changes to the safety net will depend on many factors, some of which typically lie outside the purview of a well-circumscribed empirical public policy analysis. For example, every policy

/

11

Making the Work-Based Safety Net Work Better

proposal will be scrutinized within the political system, through a lens of concerns about resources for implementation and political ramifications. Since it was beyond the scope of this volume to ask the authors to discuss the political feasibility of policy implementation and financial matters in their subject areas, we devote the final chapter of this volume to a discussion of some of these issues. In chapter 9, “The Politics of Low-Income Families in the United States,” Kent Weaver brings into focus the political challenges and constraints that have shaped not only past U.S. policy reforms to improve the economic circumstances of poor families but also continuing policy efforts to enhance the work-based safety net and increase its effectiveness. Even well-thought-out policy ideas with overwhelmingly strong research evidence undergirding them will do little to help low-income families in their quest for self-sufficiency if they cannot be enacted and implemented. The political power of low-income families is minimal, and broader public support for more generous benefits targeted to the poor appears to be weak. In his analysis of policy efforts to improve the life chances of low-income families, Weaver discusses the political dynamics of key policy “streams,” including social insurance, refundable tax credits, and cash and in-kind means-tested assistance. He identifies enduring political constraints—public opinion, policymaking bodies prone to gridlock, increasing fiscal pressures, and federalism—and considers how they might be overcome to promote more innovative, generous work-based programs for low-income families. Weaver describes the tax-based cash-transfer programs that are criticized by Meyers and Gornick as a major political success story. He points out, for example, that refundable child tax credits have become a major source of income transfers to low-income working families. He is, alternatively, pessimistic about the political chances of increasing the use of social insurance mechanisms for expanding benefits such as family leave or allowing for more generous child support, absent a pro-poor political “tsunami” or major window of opportunity. In regard to a possible broadening of Unemployment Insurance, he suggests that concerns about increased employment insecurity could allow for an incremental expansion, although given that the incidence of prolonged unemployment is low and concentrated among the poor and politically disengaged, he does not see this as likely. In general, Weaver sees that policymaking for low-income families will continue to be of relatively low political salience, although he suggests that this should not discourage advocates from aiming high and moving incrementally toward policies that promote greater self-sufficiency among the poor.

TOPICS NOT DISCUSSED IN THIS VOLUME There are some issues that would be natural additions to our volume but for a variety of reasons do not receive prominent treatment. For example, perhaps one of the most obvious topics would be adult education and skills training. In the 1970s public investments in this area were fairly substantial (over $30 billion in today’s dollars), but they have steadily been decreasing in real terms to the point where

12

/

Making the Work-Based Safety Net Work Better

they have declined by over 80 percent in constant dollars in proportion to the size of the economy. Nonetheless, any efforts to enhance skill formation will undoubtedly play a substantial role in helping low-skilled individuals to achieve self-sufficiency. In Reshaping the American Workforce in a Changing Economy, edited by Harry Holzer and Demetra Smith Nightingale (2007), contributors thoroughly discuss the issues of schooling and training for adults, including strategies for helping the hard-to-employ and increasing the labor-force participation of older workers. We refer our readers to this fine collection of writings on this topic. Scholars have long been concerned about mismatches between the geographical areas where poor people live and the jobs that might be available to these workers. Moreover, many are concerned about concentrations of low-income families and individuals, raising the possibility that individuals in poor neighborhoods may also have limited access to retail stores, attend poor-quality schools, and endure disproportionate risk for crime and gang activity. In response to these concerns, ambitious experiments with housing programs have been implemented to aid families in moving to better neighborhoods and to evaluate the role of neighborhood effects in their subsequent labor-market and quality-of-life outcomes. The resulting studies—the most prominent being a HUD-sponsored evaluation, Moving to Opportunity—have produced surprisingly modest findings. As reported in Jeffrey Kling, Jeffrey B. Liebman, and Lawrence F. Katz (2007), the intervention clearly affected the neighborhoods that people lived in; four to seven years after random assignment, families that received vouchers to move out of their poor neighborhood lived in safer, lower-poverty neighborhoods than households that did not receive vouchers. However, there were no significant effects of the intervention on adult economic self-sufficiency (employment and earnings) or physical health.16 Given this and other extensive evidence from the high-quality experimental intervention (see, for example, “Recent Moving to Opportunity Research,” available at http://www.nber.org/~kling/mto/recent.html), we chose not to include the spatial mismatch of housing and jobs as a topic for this volume. In the face of rising income and asset inequality in the United States, growing attention is being paid to issues surrounding wealth accumulation and asset development for low-income families. The federal Assets for Independence program, for example, provides funding to community-based nonprofits and government agencies that give low-income families an opportunity to save their earned income in special matched savings accounts called Individual Development Accounts (IDAs) for the purpose of acquiring a first home, starting a small business, or enrolling in postsecondary education or training. In fact, the stated rationale for IDAs is that whereas welfare and other cash transfers will reduce hardships faced by the poor, cash transfers do not help low-income families become economically self-sufficient. Gary Engelhardt et al. (2008) evaluate the results of a controlled field experiment implemented in Tulsa, Oklahoma, that examined the effects of IDAs. The authors conclude that “despite strong incentives, regular interaction between program staff and treatment group participants, and the presence of a strongly motivated group of savers, we find generally weak sample-wide effects of the Tulsa

/

13

Making the Work-Based Safety Net Work Better

IDA program on household behavior. There are no sample-wide impacts on holdings of subsidized assets” (1525). There is a significant increase in the number of renters who become home owners, but the increase in housing wealth is largely offset by a reduction in nonretirement financial assets. Moreover, the experiment included incentives for renters to accelerate home purchases and for control group members to delay home purchases.17 Although we concur that an interesting set of issues and opportunities is presented by these interventions and the corresponding policy discussions, in our opinion the benefits of these programs are considerably less likely to be as farreaching as programs focused on early-childhood development, health, education, and labor-force participation (topics covered in this volume).18 Jay Bhattacharya and Peter Richmond note (see chapter 3) that health insurance coverage is an important factor in explaining gaps in health outcomes between poor workers and nonworkers, and thus health insurance issues represent an important omission from our volume. Yet the literature on reforming health care in the United States is voluminous and growing, and arguably too expansive to be effectively integrated into the health chapter in this volume, or maybe even to be adequately addressed in a separate chapter. In reviewing the widely dispersed research on this topic, we found much of the important new work on health insurance coverage relevant to our themes of this volume to be accessible in the archives of the Economic Research Initiative on the Uninsured (http://eriu.sph. umich.edu). For example, a recent study by Thomas DeLeire, Judith A. Levine, and Helen Levy (2006) finds that less-skilled women are twice as likely to be uninsured as women who complete high school, and more than three times more likely to lack coverage than female college graduates. Acknowledging the complexity of these issues, however, they also report that welfare reform had a small positive effect on coverage trends for less-skilled women who had not received welfare benefits prior to reform. In related research, Hanns Kuttner and Catherine McLaughlin (2006) show that, among uninsured adults, job changes are the leading reason that adults become uninsured, which supports our substantial attention to labor-market policies in this volume. Kent Weaver also briefly takes up the politics of health-care reform in chapter 9. A topic of fundamental importance when considering the problems of lowincome families and individuals is family formation and fertility, for family structure has a critical role in determining whether earnings, augmented with public benefits, are sufficient for meeting basic needs. This volume’s focus on work directs attention to individual earnings, but the challenges of being self-sufficient are presumably greater for households with a single adult and dependents than it is for households with two adults and children. A recent analysis of data from the Urban Institute’s National Survey of America’s Families confirmed that being in a married two-parent family reduces risk of hardship, regardless of the family’s immigration status, race, education level, and the ages of family members. Indeed, in light of facts such as this, one of the original goals of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 was to promote marriage. Thus, we recognize that work, self-sufficiency, marriage, and fertility decisions

14

/

Making the Work-Based Safety Net Work Better

are clearly intertwined. For excellent treatments of this and related topics, see Maria Cancian and Deborah Reed (forthcoming), and the numerous works by Urban Institute researchers such as Robert Lerman, Greg Acs, and others (see http://www.urban.org/toolkit/issues/marriage.cfm). Although Kent Weaver gives some attention to the larger policy implementation challenges in chapter 9, particularly those concerning fiscal capacity and institutional fragmentation, adequate attention to the implementation issues in each of the safety-net areas discussed in the chapters could be a book in itself. Our strategy here is to present promising policies that we expect could reasonably be implemented with available resources. We understand that important choices about program design and implementation have to be made and that these choices will be different for each of these areas. Policy reforms could vary, for example, in whether they are aggressive or incremental; are conducted nationally or left to state-by-state options; include private sector participants or are the sole responsibility of public agencies. In fact, we will be delighted if some of the policies discussed in this book get far enough along in the political and policy process that their implementation becomes a topic of serious concern by policymakers and interested parties.

CROSS-CUTTING THEMES AND QUESTIONS The issues raised in the chapters that follow provide a rich set of observations in their respective domains of interest. But one of the inevitable challenges in policy design and implementation—particularly in considering the expansive scope of the social safety net—is to step outside one’s own area of interest to see how the pieces fit together. For example, a scholar or policymaker who spends most of his or her time thinking about the employment of and hours worked by disadvantaged workers may advocate policy reforms that would increase employment (akin to the increased work participation requirements in the reauthorization of PRWORA). But if employment has harmful effects on children or on the health of parents, the ultimate success of those policies will likely be jeopardized. Moreover, employment policies that target one set of individuals but that fail to account for empirically numerous subpopulations, such as ex-offenders or individuals who because of cognitive limitations, drug or alcohol problems, or other barriers are unable to comply with rules, will be unsuccessful. Thus, to increase the chances for successful policy changes, we need to look across safety-net domains to design innovative policies and reconcile high-quality research evidence in ways that will enhance the life chances of parents with low skills and the children being raised in their families and improve the communities they live in. We think the historical evidence makes it clear that a work-based safety net is viable. The striking increase in women’s labor-force participation over the past two decades, even for poor women with low levels of education and those who are unmarried with children, show that work is possible for individuals who historically have had low employment rates. Moreover, it is a political necessity that the safety net reflect the values and expectations of broader society.

/

15

Making the Work-Based Safety Net Work Better

When rates of female labor participation across all ages and education levels are high, it seems appropriate that the safety net be organized around the same values and behavior that are adopted across the population. Based on evidence from the pre-1996 period of welfare waivers and from evaluations associated with welfare reform, there is a great deal of evidence about policy levers that can increase the employment of low-skilled workers. If increasing employment is the sole goal, this goal can be achieved by strictly enforced work requirements, perhaps coupled with time-limited assistance. Alternatively, approaches that subsidize employment, raising the returns to work, have also been successful in increasing employment. The combination of work requirements, earnings subsidies through the earned income tax credit, coupled with administrative changes, perhaps reinforced by the message of time limits, appears to be a successful mix. Duncan, Gennetian, and Morris argue in “Parental Pathways to Self-Sufficiency and the Well-Being of Younger Children” (chapter 4), however, that reform packages that increase family income, generally through earnings subsidies, are superior in their effects on children to work-increasing reform packages that do not increase family income. Hence, when one broadens perspective about policy goals, employment and child well-being can be enhanced by workoriented policies that raise family income. Of course this combination of policies is more expensive to taxpayers, at least in the short run. The research presented in this volume also makes clear that a serious workoriented safety net cannot neglect the astonishing educational and performance gaps between poor and rich, and black and white students in K-through-twelve schools. This, too, is a case in which it may be easier to focus on an isolated component of the problem, such as test score achievement gaps, but a comprehensive policy must simultaneously consider linkages between child development, primary and secondary schooling, and labor markets. If children from poor families are not coming to school prepared to learn, and if schools are ineffective in helping them to overcome their disadvantage, they will inevitably have systematically worse life chances than children from other families. These children will struggle as adults and, like their parents, have difficulty fulfilling their obligation to work. Given this current state of affairs, it seems clear that if we maintain a safety net that emphasizes work, we must make it a priority to increase family incomes, facilitate access to affordable high-quality child care, and simultaneously adopt (and evaluate) hard-headed, evidence-based policies that improve teacher and school quality, particularly in schools attended by poor children. In fact, this is almost precisely the point made by Meyers and Gornick (chapter 8) when they argue that the benefits available through the safety net should be more broadly defined to include social, educational, health, and employment provisions that relate directly to the challenges facing low-earning adults who are struggling to meet the demands of working and caring for children. Their suggested reforms also go further than most others in this volume toward recognizing that periods of employment instability or unemployment are going to be unavoidable for many of the low-income groups with whom we are concerned here, and that a continued emphasis on benefits that are tied to work, through the tax system and employers, will be detrimental to these families and particularly the children who live in them. 16

/

Making the Work-Based Safety Net Work Better

But Weaver’s analysis of the “politics of permanent austerity” and the lack of broader political interest in reform of policies toward the poor suggests that the choices ahead, as efforts are taken to improve the safety net, will be difficult and constrained. One dimension of this policy tension will surely be generational. As reform efforts are undertaken, should attention be focused on interventions that come earlier in life versus those that come later for disconnected individuals, the formerly incarcerated, high school dropouts, and others? The work described in the following chapters indirectly raises this concern but does not provide detailed guidance on this important question. As Weaver’s chapter notes, policies focused on children have over time received broader political support than policies focused on adults or other groups. There is also a considerable amount of evaluation evidence suggesting that well-funded early-childhood interventions can strikingly increase the school performance, behavior, graduation, and even labormarket outcomes of disadvantaged youth at a reasonable cost, that is, with high rates of return.19 Thus, in light of the strong evidentiary basis for investing in welldesigned early-childhood programs, should we invest aggressively in these programs, acknowledging that this would inevitably limit our ability to make progress in other policy areas important to older youths and adults? Children, of course, are raised in households with adults. The training literature suggests that some in-school youth training programs show promising results, but recent programs for out-of-school youths have not been evaluated. Results for older programs targeting those sixteen to twenty-four have generally shown poor results. Adult training programs, particularly for low-skilled women, on the whole have been more successful in increasing workers’ earnings. One likely reason is that adults (those older than twenty-four) have had sufficient life experience to take more seriously the opportunities provided by training. In addition, employment and training programs have evolved over time to more effectively meet both employer and worker needs, including better targeting of local employment sectors that are growing rapidly and pay more favorably, and making better use of financial incentives that support individuals’ transitions to and retention of employment. Recent research by Harry Holzer and others in the volume edited by Holzer and Nightingale (2007) suggests that there is substantial untapped potential for improving and expanding training options for low-skilled adults.20 Thus, it seems to us that it would be a mistake to forego these types of opportunities for those of any age who wish to improve their ability to be selfsufficient and contribute more to their communities. So even though the existing evidence in support of early-childhood programs is strong, we feel that a balanced approach to enhancing self-sufficiency across age groups is wise. An equally difficult question is whether the current system, which yields extraordinary differences in outcomes between rich and poor, black and white, and urban core versus suburban, can be improved with shorter-term, incremental policy fixes, or whether bolder, longer-term reforms will be necessary to effectively increase self-sufficiency. There is no easy answer to this question. Our idiosyncratic view is that safety-net policy developments in the past twenty years have had many valuable features, including the realignment of the central message of the safety net—that work is at the basis of economic well-being—with the core /

17

Making the Work-Based Safety Net Work Better

values of American society. Within that realignment, many innovations have in our view been steps in the right direction, including expansions of the Earned Income Tax Credit, State Children’s Health Insurance program (S-CHIP) and Children’s Health Insurance Program (CHIP) expansions that sharply increased health insurance coverage of children, and child support and welfare reforms that sought to increase the resources available to low-income families for caring for their children, such as child support pass-through, child-care assistance, and others. Nevertheless, the problems confronting low-income families in the United States are in some cases so severe that trying to solve them using what resources are being devoted to them is akin to trying to slay an elephant with a pea shooter. For example, we say little in this volume about complex and intractable problems such as substance abuse and mental health problems that affect a significant number of poor families and undoubtedly contribute greatly to their disadvantage. Still, we think the agenda described in the following chapters, augmented with a more serious attempt at comprehensively providing health care for low-income children and adults, could usefully guide the expenditure of greater resources than we are currently spending. We would probably characterize ourselves as incrementalists, since we think the history of success in antipoverty policy has often come through the enactment of initially small programs, or through expansions of existing programs. At the same time, we welcome “bigger bangs” that would enact evidence-based reforms. In either case, we both would devote considerably more of society’s resources to these issues. So what are some examples of these “bigger bangs”? First, we suggest that when the evidence base is very strong in support of the cost-effectiveness of a particular intervention, this is a clear signal to move more aggressively. For example, it seems extraordinary that we have failed to invest more in early-childhood education, given that the evaluation evidence suggests the social returns to these investments are substantially greater dollar for dollar than for other types of programs. More generally, the reduced costs of crime, teen childbearing, and the additional resources society would eventually receive in higher taxes and productivity from greater educational attainment would more than pay for the costs of providing high quality early-childhood programs and more effective schools. We should seize opportunities for policy improvement in these areas, including those suggested in this volume, which we expect will also improve the material circumstances of disadvantaged children. Second, we think that Steven Raphael’s work identifies a substantive policy area, criminal justice, in which policy reform efforts have failed miserably over the last few decades, with devastating consequences for low-income families and communities. Of the more than $40 billion spent annually on prison and jail costs, over half goes toward the incarceration of nonviolent offenders, and evidence confirms that recent spending increases on incarceration have been less effective in reducing crime than expenditures on sensible alternatives, such as early childhood development. Raphael also shows how current state and federal policies have compounded the problems of former inmates as they attempt to secure employment and reenter noninstitutionalized society. Given the collateral consequences on the family of sharply impairing the employment prospects of the for18

/

Making the Work-Based Safety Net Work Better

merly incarcerated and the lack of effective policy programs targeted to this problem to date, it seems particularly urgent to embark aggressively on a series of well-designed experimental projects and corresponding evaluations to develop the evidence base for approaches that can successfully achieve labor-market success for ex-offenders and improved family and community well-being. A final difficult question arises over the appropriate locus of program responsibility. Welfare reform in the United States aggressively devolved programmatic responsibility for welfare from the federal government to states. The justification came in many pieces. Two prominent ones were that states and localities are more familiar with the specific circumstances of the people in their communities and hence are better able to design assistance programs; and that states and localities can serve as laboratories, where innovation can flourish and other communities can learn from best practices. A problem with the second justification is that at best weak evaluation requirements were associated with devolution. It is very hard to learn what works and what does not when there is little systematic evaluation of new policy efforts and experiences. Nevertheless, the question of whether policies are best designed and implemented at the federal or at the state level raises important issues. Our biases are that federal involvement—not necessarily centralized control—is necessary for at least three reasons: financing, standards, and dissemination. The federal government is less susceptible to regional economic shocks and thus better able to provide a stable funding stream for effective programs. Moreover, in return for providing the financial base, the federal government can impose data collection and evaluation standards on states. Absent this, it is nearly impossible to learn about successful programs. With serious program evaluation requirements in place, the federal government is in a position to disseminate information on policy design, implementation, and outcomes and to promote the adoption and diffusion of successful, evidence-based programs. We think the thoughtful, forward-looking recommendations presented in this volume, based on the rigorous social science research of the contributors, provide an excellent starting point for renewed policy efforts to strengthen the work-based safety net and enhance the well-being and self-sufficiency of low-income families.

NOTES 1. These figures, the civilian labor force participation rates in 2007, come from the Bureau of Labor Statistics table generation program. Available at: http://data.bls.gov/ PDQ/outside.jsp?survey=ln (accessed on April 14, 2008). 2. See H.R. 3734, section 401. Available at: http://thomas.loc.gov/cgi-bin/query/ z?c104:H.R.3734.ENR: (accessed March 23, 2009). 3. In Wisconsin, for example, nearly two-thirds of welfare recipients had earned less than $2,500 in the two years prior to when they were observed on welfare in July 1995, and only 17 percent had earned more than $7,500 (Cancian et al. 1999). These families had on average more than two children. 4. See, for example, Haveman and Scholz (1994). 5. See, for example, Ziliak (2002). /

19

Making the Work-Based Safety Net Work Better 6. Mead (2007) writes that low earnings among poor, low-skilled adult men are not “due principally to low pay but rather to a failure to work consistently at any job.” He further argues that part of the failure of past programs is that they have given men a “choice” to work rather than making it an “obligation.” See also Haskins and Sawhill (2007). 7. In 1995, the poverty line for a two-person (parent and child) family was $10,030; $12,590 for a three-person family; and $15,150 for a family of four. In 1997 the poverty line for a two-person family was $10,610; $13,630 for a three-person family; and $16,050 for a family of four. The figures were even less favorable for those who left welfare in the fourth quarter of 1997: average earnings in the first year were $7,709 (the median was $6,662). The lower earnings of those in the second group of leavers were not necessarily surprising, given their greater disadvantage relative to those leaving in 1995. They were clearly less educated: only 45.6 percent of recipients in 1997 had at least a high school diploma, compared with 56.4 percent in 1995. 8. Wisconsin’s unemployment rate in 1995 was 3.7 percent, compared to the national rate of 5.6 percent; in 1999 it was 3.0 percent, against a national rate of 4.2 percent. Wisconsin’s welfare reform program, Wisconsin Works (called W-2), was implemented well before other programs, so that the early Wisconsin experiences may be a harbinger of what families in other states may experience. Moreover, W-2 places great emphasis on work; to make work feasible it provides more generous support for health insurance and child care than most other states. With aggressively work-oriented programs in place and a very strong state economy, the labor-market experiences of people leaving welfare in Wisconsin are likely to be as good as one would find anywhere. 9. LaDonna Pavetti and Gregory Acs (2001) also find little evidence that low-skilled women with children will easily move into jobs that allow them to have incomes above the poverty line, as they acquire additional labor-market experience. 10. Even more troubling, the data used to examine the labor-market fortunes of individuals leaving welfare are drawn from a period when the economy was very strong, and at least in Wisconsin there were significant state expenditures for work-supporting child-care and health insurance programs. 11. One exception is the 2004 report by MDRC and the National Governor’s Association, “Building Bridges to Self-Sufficiency: Improving Services for Low-Income Working Families,” which sought to identify “promising practices” and innovative programs under way in states and localities to more effectively support low-income, working populations and their families. See http://www.mdrc.org/publications/385/full.pdf (accessed August 14, 2008). 12. David Neumark’s conclusions about the minimum wage are somewhat controversial (see chapter 2). 13. Serum markers are substances that are soluble in the serum (noncellular portion of blood) that are present at high levels in association with specific diseases. 14. It is not yet clear from the research evidence whether this is because the policies are ineffective or because these factors do not matter very much. 15. Fewer than 15 percent of poor U.S. children live with nonworking parents. 16. Female youths in the treatment group did better on measures of education, risky behavior, and physical health, but male youths did worse on each dimension. The experiment did reveal mental health benefits for all adults and female youths. 17. A more optimistic view of IDAs is given in a nonexperimental longitudinal study. It 20

/

Making the Work-Based Safety Net Work Better concluded that after three years, participants were significantly more likely to be homeowners and business owners and to engage in postsecondary education than nonparticipants. The comparison group of nonparticipants was selected from the 2001 Survey of Income and Program Participation, and propensity score matching was used to adjust for observed differences between IDA participants and nonparticipants. See Abt Associates, Inc. (2008). 18. A forthcoming volume edited by Michael Barr and Becky Blank will provide a thorough discussion of these issues and their implications for financial services policy. 19. There is, of course, more that could usefully be learned about the difficulties that arise in bringing small programs to larger scale, and about the speed at which program benefits may fade out. 20. For a thorough discussion of the range of federal, state, and local employment training programs and some of the most promising approaches among them, see Holzer (2007.

REFERENCES Abt Associates, Inc. 2008. “Assets for Independence Act Evaluation Impact Study: Final Report.” Prepared for the Administration for Children and Families, U.S. Department of Health and Human Services, Washington, D.C. Available at: http://www.acf.hhs.gov/pro grams/ocs/afi/AFI_Final_Impact_Report.pdf (accessed March 23, 2009). Cancian, Maria, Robert Haveman, Thomas Kaplan, and Barbara Wolfe. 1999. “Post-Exit Earnings and Benefit Receipt Among Those Who Left AFDC in Wisconsin.” IRP Special Report 75. Madison: University of Wisconsin, January. Available at: http://www.irp .wisc.edu/publications/sr/pdfs/sr75.pdf (accessed March 23, 2009). Cancian, Maria, Robert Haveman, Daniel Meyer, and Barbara Wolfe. 2000. “Before and After TANF: The Economic Well-Being of Women Leaving Welfare.” IRP Special Report 77. Madison: University of Wisconsin, May. Available at: http://www.irp.wisc.edu/publica tions/sr/pdfs/sr77.pdf (accessed March 23, 2009). Cancian, Maria, and Deborah Reed. Forthcoming. “Changes in Family Structure, Childbearing, and Employment: Implications for the Level and Trend in Poverty.” In Changing Poverty, edited by Maria Cancian and Sheldon Danziger. New York: Russell Sage Foundation. Citro, Constance, and Robert Michael. 1995. Measuring Poverty: A New Approach. Washington, D.C.: National Academy Press. DeLeire, Thomas, Judith A. Levine, and Helen Levy. 2006. “Is Welfare Reform Responsible for Low-Skilled Women’s Declining Health Insurance Coverage in the 1990s?” Journal of Human Resources 41(3): 495–528. Ellwood, David T. 1988. Poor Support: Poverty in the American Family. New York: Basic Books. Engelhardt, Gary, Gregory Mills, William G. Gale, Rhiannon Patterson, Michael Erikson, and Emil Apostolov. 2008. “Effects of Individual Development Account on Asset Purchases and Saving Behavior: Evidence from a Controlled Experiment.” Journal of Public Economics 92(5–6, June 2008): 1509–30. Gardiner, Jean. 2000. “Rethinking Self-Sufficiency: Employment, Families and Welfare.” Cambridge Journal of Economics 24(6): 671–89. Garfinkel, Irwin, and Sara McLanahan. 1986. Single Mothers and Their Children: A New American Dilemma. Washington, D.C.: Urban Institute Press. /

21

Making the Work-Based Safety Net Work Better Gladden, Tricia, and Christopher Taber. 2000. “Wage Progression Among Less Skilled Workers.” In Finding Jobs: Work and Welfare Reform, edited by D. Card and R. Blank. New York: Russell Sage Foundation. Grogger, Jeffrey, and Lynn A. Karoly. 2005. Welfare Reform: Effects of a Decade of Change. Cambridge, Mass.: Harvard University Press. Gueron, Judith M. 1990. “Work and Welfare: Lessons on Employment Programs.” Journal of Economic Perspectives 4(1): 79–98. Haskins, Ron, and Isabel Sawhill. 2007. “The Next Generation of Antipoverty Policies.” Special issue, The Future of Children 17(2), Fall. Haveman, Robert, and John Karl Scholz. 1994. “Transfers, Taxes, and Welfare Reform.” National Tax Journal 47 (June): 417–34. Holzer, Harry, and Demetra Smithe Nightingale, eds. 2007. Reshaping the American Workforce in a Changing Economy. Washington, D.C.: Urban Institute Press. Hotz, V. Joseph, Guido Imbens, and Jacob Klerman. 2006. “Evaluating the Differential Effects of Alternative Welfare-to-Work Training Components: A Re-Analysis of the California GAIN Program.” Journal of Labor Economics 24(3, July): 521–66. Hotz, V. Joseph, Charles Mullin, and John Karl Scholz. 2002. “Welfare Benefits, Employment and Income: Evidence from the California Work Pays Demonstration Project.” Available at: http://www.ssc.wisc.edu/~scholz/Research/CWPDP_paper_version6.pdf (accessed March 23, 2009). King, Christopher T., and Peter R. Mueser. 2005. Welfare and Work Experiences in Six Cities. Kalamazoo, Mich.: Upjohn Institute for Employment Research. Kling, Jeffrey R., Jeffrey B. Liebman, and Lawrence F. Katz. 2007. “Experimental Analysis of Neighborhood Effects.” Econometrica 75(1): 83–119. Kuttner, Hanns, and Catherine McLaughlin. 2006. “Job Churning’s Impact on Workplace Health Insurance.” ERIU Research Highlight no. 12. Ann Arbor, Mich.: Economic Research Initiative on the Uninsured, November. MDRC. 1994. Executive Summary of GAIN: Benefits, Costs, and Three-Year Impacts. New York: June. Mead, Lawrence M. 2004. Government Matters: Welfare Reform in Wisconsin. Princeton, N.J.: Princeton University Press. ———. 2007. “Toward a Mandatory Work Policy for Men.” In “The Next Generation of Antipoverty Policies.” The Future of Children (special issue, edited by Ron Haskins and Isabel Sawhill) 17(2, Fall): 43–72. Meyer, Daniel, and Maria Cancian. 1998. “Economic Well-Being Following an Exit from Aid to Families with Dependent Children.” Journal of Marriage and the Family 60(2): 479–92. Pavetti, LaDonna, and Gregory Acs. 2001. “Moving Up, Moving Out, or Going Nowhere? A Study of Employment Patterns of Young Women and the Implications for Welfare Mothers.” Journal of Policy Analysis and Management 20(4): 721–36. Primus, Wendell, Lynette Rawlings, Kathy Larin, and Kathryn Porter. 1999. The Initial Impacts of Welfare Reform on the Incomes of Single-Mother Families. Washington, D.C.: Center for Budget and Policy Priorities, August. Ziliak, James P. 2002. “Social Policy and the Macroeconomy: What Drives Welfare Caseloads?,” Focus 22(1): 29–34, Special Issue on Reauthorizing TANF 2002, Institute for Research on Poverty, University of Wisconsin, Madison.

22

/

Chapter 2 Alternative Labor-Market Policies to Increase Economic Self-Sufficiency: Mandating Higher Wages, Subsidizing Employment, and Increasing Productivity David Neumark

T

he principal means by which individuals and families achieve economic self-sufficiency is through labor-market earnings. As a consequence, it is natural for policymakers to look to interventions that increase the ability of individuals and families to achieve an adequate standard of living from participating in the labor market. This chapter discusses some key policies that are used or can be used to increase economic self-sufficiency by increasing earnings. Broadly speaking, the set of available policy options can be cast in the context of a simple supply-and-demand analysis. First, we can try to increase earnings via higher wages, by mandating a higher wage floor. Second, we can try to increase earnings by subsidizing employment on either the demand or the supply side, either raising the demand for labor or increasing the supply of labor. And third, we can encourage greater human capital accumulation, increasing productivity and thereby shifting the demand curve out and raising wages and employment.1 Many public policies directed at the labor market can be viewed within this framework. In this chapter I emphasize policies on which I can bring my own expertise to bear and for which I can provide what I view as the most valuable contribution to the policy debate by highlighting recent research. In particular, the chapter focuses on mandated wage floors (minimum and living wages), employment incentives (the Earned Income Tax Credit and wage subsidies), and school-to-work policies. Obviously, the potential scope of this chapter—what government policy can do to increase workers’ and families’ earnings—is far more extensive.2 Innumerable papers and many books have been written about welfare

/

25

Making the Work-Based Safety Net Work Better

reform, job training, the EITC, the minimum wage, and educational reform. I can neither touch on all of the potential policy levers nor even fully treat those policies I do discuss.3 Policymakers have attempted to increase earnings by mandating higher wages via minimum wages and, more recently, living wages. Of course, mandating higher wage floors has the potential to discourage employment of low-skill workers, and this is the source of much of the controversy over minimum-wage floors. Perhaps the most prominent set of policy interventions and changes with respect to increasing earnings in pursuit of economic self-sufficiency have targeted the supply side, by trying to change the incentives to work. Welfare reform has clearly aimed to increase employment of the target population—low-income households and low-skilled single mothers in particular—as have other policy changes, including expansions of the Earned Income Tax Credit (EITC), revised income tax schedules, and modifications to Medicaid and the provision of public health insurance to children. These policy interventions have sought to change labor supply on either the extensive or the intensive margin. There seems to be little doubt that these policy interventions have contributed to higher employment among the target population (for example, see Meyer and Rosenbaum 2000; Blank and Schmidt 2001; Blank 2002). Viewed from the perspective of increased employment and reduced caseloads, the combined effects of welfare reform and these other changes appear to have been a successful effort to increase earnings— although that, of course, does not mean families are better off.4 An alternative policy that operates instead on the demand side is a wage subsidy program targeted toward low-skilled or disadvantaged individuals. Wage subsidy programs can take many forms, depending in part on who is targeted, but all share the basic structure of subsidizing wages to increase demand for workers (shifting out the labor demand curve), thereby raising their employment and earnings (see Katz 1998). The current incarnation of wage subsidies in the United States is the Work Opportunity Tax Credit (WOTC), which targets young workers in disadvantaged families or who are “high risk,” and members of families receiving TANF (Temporary Assistance for Needy Families), as well as a few other groups.5 Of course, a key alternative to increasing incentives for work and increasing wages or demand for low-skilled or disadvantaged workers is to try to directly raise the productivity of such workers through training and education. Looking through this lens, one could view much of the entire body of research on human capital as pointing to ways to increase earnings. For example, there is little disagreement that schooling increases earnings, even if there is continuing debate about the magnitude of the effect. And a simple policy goal of trying to increase schooling is even more compelling in light of increased returns to schooling in recent decades.6 In addition, one could view research by labor economists linking educational reform to earnings (for example, Betts 1995; Card and Krueger 1992; Hanushek 2006) as also pointing to policy interventions to increase economic selfsufficiency via earnings. I focus more narrowly on one dimension of schooling- and training-related

26

/

Alternative Labor-Market Policies

policies for which policy experience is quite recent and for which research findings have only recently begun to emerge—in particular, school-to-work. This focus is not intended to suggest anything about the weighting of the importance of alternative human capital–related policies for increasing earnings, but rather to add information on what we have learned from the recent experience with and research about school-to-work to the broader research literature on human capital and educational reform.

MANDATING HIGHER WAGES The minimum wage has been a central component of the nation’s efforts to help families achieve economic self-sufficiency since early in the last century. The federal minimum was instituted in 1938, and was predated by earlier efforts in some states. Until the early 1980s, increases in the federal minimum wage were sporadic but over the longer term largely preserved the real value of the minimum wage. Since then, federal increases have slowed, and instead states have picked up the ball; as of January 2007, twenty-nine states plus the District of Columbia had minimum wages above the federal level.7 A related development has been the advent of living wages since the mid-1990s, which have now spread to over 140 cities and other localities. Do minimum wages and living wages enable individuals to earn more and families to exit and remain out of poverty? The research literature is enormous. William Wascher and I recently completed an extensive review of the evidence on the employment effects of minimum wages (Neumark and Wascher 2007) and a broader survey of the effects of minimum wages and living wages on a variety of outcomes (Neumark and Wascher 2008). Here, I discuss some of the main conclusions.

Minimum Wages and Employment Much of the political debate surrounding proposed changes in the minimumwage concerns the potential effects on employment. Although that focus is not entirely appropriate, the potential disemployment effects are of course the channel that could dissipate or even outweigh the gains to low-skilled individuals from higher wages. A large body of research on the employment effects of the minimum wage, conducted in the 1970s, focused on the effects of the federal minimum wage and consisted mostly of time-series studies. The most widely cited survey of this literature (Brown, Gilroy, and Kohen 1982) concluded that the evidence established a consensus range of –.1 to –.3 for the elasticity of teenage employment with respect to the minimum wage; these elasticities measure the ratio of the percent change in employment to the percent change in the federal minimum wage. The state-level increases since the late 1980s, coupled with federal increases in 1990 and 1991,

/

27

Making the Work-Based Safety Net Work Better

provided economists with additional “experiments” with which to reexamine the costs and benefits of minimum wages, and this new wave of research produced considerably more diverse results. The best-known and most controversial of these new studies were conducted by David Card and Alan Krueger (summarized in their 1995 book Myth and Measurement). They argued that the consensus view that minimum wages reduce employment of less-skilled workers was wrong, and their most widely cited results on the fast-food industry suggested that a higher minimum wage could lead to substantial increases in low-skilled employment.8 Other studies published during this period, sometimes from other countries, cast doubt on the conclusion that minimum wages reduce employment of low-skilled workers (for example, Machin and Manning 1994). This research was sometimes perceived as ending the consensus among economists that minimum wages reduce employment among the less skilled. However, the studies highlighted as challenging this consensus were part of a much larger body of research on the employment effects of minimum wages conducted since the early 1990s, including a number of studies I wrote with William Wascher. This larger body of new research includes more than one hundred studies and encompasses an impressive variety of statistical techniques and datasets, including more sophisticated time-series analyses, case studies of particular minimum-wage increases, and panel studies across states and years. Wascher and I recently undertook an exhaustive review in order to determine what can be learned from the wider body of research (Neumark and Wascher 2007). This review led to two main conclusions. First, there certainly is a much wider range of estimates of the effects of the minimum wage on employment than was the case in the earlier time-series literature reviewed by Charles Brown, Curtis Gilroy, and Andrew Kohen (1982). For example, few of the studies in the survey were outside the consensus range of –.1 to –.3 for the elasticity of teenage employment with respect to the minimum wage. In contrast, even limiting the focus to studies of the effects of the minimum wage on teenagers in the United States, the range of estimates in the more recent research extends from well below –1 to well above zero. This wider range of estimates partly reflects the fact that the newer literature uses a variety of methods and data to identify the effects of minimum wages—including estimates for narrow subsets of workers and specific industries—whereas the earlier literature was for the most part based on aggregate time-series data that changed only by the addition of more data with the passage of time (although there were advances in statistical methods). Second, despite the wider range of estimates, the oft-stated assertion that the new minimum-wage research fails to support the view that the minimum wage reduces the employment of low-skilled workers is clearly incorrect. Indeed, the preponderance of the evidence points to disemployment effects. Of 102 studies on which we focus, by our reckoning nearly two-thirds give a relatively consistent (although by no means always statistically significant) indication of negative employment effects of minimum wages, whereas only eight give a relatively consistent indication of positive employment effects. In addition, we identify thirty-

28

/

Alternative Labor-Market Policies

three studies that we view as providing the most credible evidence. Among these, twenty-eight (85 percent) point to negative employment effects. Moreover, in research that focuses on the least-skilled groups most likely to be adversely affected by minimum wages, the evidence for disemployment effects seems especially strong, with minimum wages harming the least-skilled workers more than is suggested by the net disemployment effects estimated in many studies.9 In contrast, we see very few, if any, cases where a study provides convincing evidence of positive employment effects of minimum wages.10 Overall, this review of the newer literature largely solidifies the conclusion that minimum wages reduce employment of low-skilled workers.11

The Distributional Effects of Minimum Wages Despite minimum wages reducing employment among the less skilled, minimum wages could have offsetting beneficial effects because of income gains resulting from higher mandated wages. And it is possible that these income gains tend to accrue to the lowest-income families, an argument sometimes made by those who advocate minimum-wage increases.12 Although the distributional question is central, it often receives short shrift in research and debates about the minimum wage, with critics of minimum wages ignoring the possibility that even if minimum wages do reduce employment, they may have beneficial distributional effects, and advocates simply assuming that the distributional effects must be beneficial. In fact, as we explain in this section, research fails to establish that minimum wages have beneficial distributional effects that outweigh the employment losses. It is commonly claimed that existing estimates of minimum-wage employment elasticities imply that minimum wages must, on average, raise incomes of lowwage workers. The argument is that if the employment elasticity for these workers is in the –.1 to –.3 range suggested by many studies of teenagers and young adults, then because the elasticity is below 1 in absolute value, a higher minimum wage must raise incomes of affected workers (Freeman 1996). However, this argument is likely incorrect. Teenagers or young adults are typically studied in research on the employment effects of minimum wages because a large share of them work at or near the minimum wage, so that the effects of minimum wages are more likely to be evident for these groups than for others. Nonetheless, many teenagers and young adults earn significantly more than the minimum wage. As a result, the reported elasticities from studies of teenagers will tend to understate the elasticity of demand with respect to the minimum wage for the least-skilled workers among them who are directly affected by the minimum wage. The same argument applies with greater force to the broader adult population, because the share of adult workers at the minimum is much smaller. The estimated elasticity from the usual minimum-wage study also underestimates the relevant elasticity of demand for affected workers because, with some affected workers already earning more than the old minimum wage, when the minimum wage increases,

/

29

Making the Work-Based Safety Net Work Better

the size of the average wage increase associated with a minimum-wage increase likely will be less than the minimum-wage increase itself.13 To examine directly how minimum wages affect low-wage workers, Neumark, Mark Schweitzer, and Wascher (2004) estimate various margins of minimum-wage effects, including wages, employment, hours (conditional on remaining employed), and—most important for the purposes of this discussion—earned income, using individual-level matched observations on those aged sixteen and older from the Current Population Survey (CPS) Outgoing Rotation Group (ORG) files for the years 1979 to 1997. For each outcome we estimate a model that interacts the change in the effective minimum wage for each state-month observation with a set of indicator variables that describe where each individual’s initial wage stands in relation to the minimum wage.14 With the additional controls included, minimum-wage effects are identified from differential changes in outcomes for workers at similar points in the wage distribution who experience different minimum-wage changes, and the approach generates estimates of the effects of minimum-wage increases on these outcomes at various points of the wage distribution. We estimate both contemporaneous effects and cumulative effects that allow one-year lags. The contemporaneous elasticity of wages with respect to the minimum is about .8 for workers at the minimum or below 1.1 times the minimum.15 The elasticity falls to about .4 for workers between 1.1 and 1.3 times the minimum, to about .25 for workers between 1.3 and 1.5 times the minimum, and to .15 for workers between 1.5 and 2 times the minimum, petering out higher in the wage distribution. However, the cumulative contemporaneous plus lagged effects are quite different, with the elasticity near the minimum wage falling to about .4, and declining for the cells slightly higher in the wage distribution. The smaller estimated effects once lags are included suggests that a substantial part of the wage gains caused by minimum-wage increases are “given back” in the following year, likely because employers forego the usual nominal wage increases in subsequent years for workers whose wages were increased by the minimum wage, while workers at the same position in the wage distribution in states without minimum-wage increases receive these nominal increases. For workers initially earning close to the minimum wage, the estimated employment elasticities range from about –.06 to –.15 and are sometimes statistically significant. The estimated elasticities are close to the so-called consensus range of estimated disemployment effects for teenagers, even though the latter estimates are based on workers whose wages can be well above the minimum wage. However, the hours reductions are more severe. The cumulative estimates point to statistically significant hours reductions for workers initially paid at or just above the minimum wage, with elasticities near –.3. Finally, we turn to earned income. The contemporaneous effects are positive (and significant for most cells) for workers initially earning up to twice the minimum wage. However, the cumulative effects tell a much different story. Workers initially below the minimum, at the minimum, and up to 1.1 times the minimum experience income declines. The estimated elasticity for minimum-wage workers is on the order of –.6 and is statistically significant at the 5 percent level; it is about

30

/

Alternative Labor-Market Policies

half that for workers just above the minimum wage. The reversal from the contemporaneous effects is attributable in part to the weaker wage gains once we get past the immediate effect of increases in the minimum wage. And overall, the fairly strong negative effects on earnings are in part due to the joint distribution of wage, employment, and hours effects across individuals. Overall, then, this analysis indicates that very low-wage workers are, on average, not helped by minimum-wage increases, and instead are hurt, despite the wage increases among those who stay employed. What about the effects on family incomes? Because the definition of poverty is based on family income, minimum-wage workers need not be in poor families. Edward M. Gramlich (1976), using data from the early 1970s, showed that there were many low-wage workers in nonpoor and even above-median-income families. More recent evidence (Burkhauser and Sabia 2007) echoes Gramlich’s earlier conclusions. For example, in March 2003 CPS data, 4.2 percent of all workers were in poor families, but only 13.2 percent of workers earning a wage less than $7.25 were in poor families. Alternatively, using a definition of low-wage workers based on half the average private-sector wage, in the 2003 data 46.3 percent, or nearly one-half, of low-wage workers were in families with incomes three times the poverty line or higher, whereas 24.2 percent were in poor or near-poor families.16 These numbers clearly imply that many of the potential benefits of a higher minimum wage could flow to higher-income families. The key question, however, is how the distribution of family incomes is actually affected by minimum-wage increases. Minimum wages undoubtedly create winners and losers, and neither the types of descriptive statistics just discussed, nor the fairly large number of studies that try to simulate the effects of minimum wages on the distribution of family incomes (for example, Horrigan and Mincy 1993; Card and Krueger 1995; Burkhauser, Couch, and Wittenburg 1996), describe the actual distributional effects of minimum wages and their incidence across families at different points in the income distribution. In order to describe more fully how minimum wages affect the distribution of family income relative to needs, Neumark, Schweitzer, and Wascher (2005) develop a difference-in-differences nonparametric approach, applied to matched March CPS files from 1986 to 1995. This approach yields estimates of the effects of minimum wages on the proportion of families that are poor or near-poor, and of the extent to which minimum wages push families initially near-poor into poverty, or lift initially poor families out of poverty. Moreover, the nonparametric approach yields a rich description of the effects of minimum wages on family incomes at all points of the income-to-needs distribution. The main results are displayed in figure 2.1. I do not show the entire set of density estimations that are used to infer the effects of minimum-wage increases on the distribution of income-to-needs, but rather just the final contemporaneous, lagged, and cumulative estimates of changes in the distribution for the treatment group of state-year observations with minimum-wage increases relative to the control group of observations without increases.17 The difference-in-differences estimates of the effects of contemporaneous minimum-wage increases on the in-

/

31

FIGURE 2.1

/ Estimated Effects of Minimum Wages on Distribution of Family Income-to-Needs, 1986 to 1995 A. Contemporaneous Effects Treatment-Control Income-Needs Densities

0.002 0.001 0 –0.001 –0.002 –0.003

0

1

2 3 Pure Contemporaneous Increase

4

5

4

5

4

5

B. Lagged Effects Treatment-Control Income-Needs Densities 0.002 0.001 0 –0.001 –0.002 –0.003

0

1

2 3 Pure Lagged Increase C. Combined Effects Summary Effects on Income-Needs

0.002 0.001 0 –0.001 –0.002 –0.003

0 Estimate

1

2 3 Income/Needs Upper Bound (95 Percent CI)

Lower Bound (95 Percent CI)

Source: Neumark, Schweitzer, and Wascher 2005. Note: The estimates are based on data taken from matched CPS March files for 1986 to 1995. See the text for explanation.

32

/

Alternative Labor-Market Policies

come-to-needs distribution are shown in panel A, with slight adjustment based on the need to account for the fact that, for example, some states with no contemporaneous increase had an increase in the previous year. The results indicate that the effect of contemporaneous minimum-wage increases is to reduce the proportion of families with income-to-needs between 0 and about .6, to increase the proportion with income-to-needs between .6 and 1.5, and to reduce the proportion with income-to-needs from 1.5 to about 2.7. These results are consistent with minimum wages helping the poorest families, but they also suggest that some families with initial income-to-needs in the range from 1.5 to about 2.7 experience income losses. Panel B of figure 2.1 reports the difference-in-differences estimates of the lagged minimum-wage effect. In contrast to the estimated effects of contemporaneous minimum-wage increases, lagged increases unambiguously raise the proportion of families below about 1.3 times the poverty line, with corresponding decreases in the proportion of families with income-to-needs between 1.3 and 3.2. This evidence, and the contrast with contemporaneous effects, is consistent with disemployment effects (or hours reductions) occurring with a lag, while the contemporaneous effect reflects more of the impact of immediate wage increases—which diminish fairly quickly, according to the worker-based results discussed earlier. The total effects of minimum-wage increases, shown in panel C, are the sums of the contemporaneous and lagged effects. The estimated effect at each particular point of the income-to-needs distribution is given by the middle curve, and the upper and lower curves are the boundaries of the 95 percent confidence interval, calculated using a bootstrap procedure. The results are quite striking. There is essentially no change in the proportion of families with income-to-needs below .3, as the benefit associated with the contemporaneous increase is offset by the cost of the lagged increase. There is a marked increase in the proportion of families with income-to-needs between about .3 and 1.4, and a marked decrease in the proportion of families with income-to-needs between about 1.4 and 3.3. These results suggest that the overall effect of minimum-wage increases is to push some families that are initially low-income but above the near-poverty line into poverty or near poverty. On a point-by-point basis, the estimated increases in the proportions of families with income-to-needs from about .6 to 1.2 are statistically significant. By integrating under the curves in figure 2.1, and bootstrapping, we find—as reported in the first row of table 2.1—that the minimum wage has essentially no effect on the proportion of families with income-to-needs between 0 and .5 but leads to significant increases in the proportion of families between .5 and 1 and the proportion below 1. There is also a significant (at the 10 percent level) increase in the proportion of near-poor families, and a statistically significant increase in the proportion of poor or near-poor families. The estimated elasticity of changes in the proportion of families that are poor or near-poor with respect to the minimum wage is approximately .41, and the average minimum-wage increase in the sample increases the proportion of families in these categories (combined) by .013, and the proportion of poor by .008. The estimates in the second row are based on

/

33

0.0079*** (0.0025) 0.0069** (0.0028)

0.5 to 1 (2) 0.0083** (0.0035) 0.0071* (0.0039)

0.0046* (0.0027) 0.0033 (0.0034)

1 to 1.5, Near-Poor (4) 0.0130*** (0.0040) 0.0104** (0.0046)

0 to 1.5, Poor/ Near-Poor (5)

–0.0049* (0.0028) –0.0072** (0.0033)

1.5 to 2 (6)

–0.0071** (0.0031) –0.0074** (0.0037)

2 to 3 (7)

–0.0120*** (0.0040) –0.0146*** (0.0048)

1.5 to 3 (8)

Source: Neumark, Schweitzer, and Wascher 2005. Notes: The data come from matched CPS March files, from 1986 to 1995. Estimates are constructed by integrating under the densities like those reported in figure 2.1. The total sample size for the analysis, including families with income-to-needs up to 6, is 196,270. Standard errors are bootstrapped, based on five hundred repetitions, with implied t-statistics asymptotically normally distributed. * p ≤ .10; ** p ≤ .05; *** p < .01

Fixed state and year effects (proportional shifts)

0.0005 (0.0018) 0.0002 (0.0022)

0 to 0.5 (1)

0 to 1, in Poverty (3)

Income-to-Needs Ranges

/ Estimated Effects of Minimum-Wage Increases on Proportions in Income-to-Needs Ranges

Changes in proportions No controls

TABLE 2.1

Alternative Labor-Market Policies TABLE 2.2

/

Wages and Family Income-to-Needs

Income-to-Needs Ranges 0 to 0.5

0.5 to 1

1 to 1.5

1.5 to 2

2 to 3

(1)

(2)

(3)

(4)

(5)

A. Distributions of primary earners in family income-to-needs category by hourly earnings Less than 90 percent of minimum 90 to 110 percent of minimum 110 to 200 percent of minimum More than 200 percent of minimum

0.49 0.17 0.25 0.09

0.27 0.18 0.43 0.12

0.12 0.12 0.53 0.23

0.06 0.05 0.50 0.39

0.03 0.02 0.29 0.66

B. Distributions of lowest earner in family in family income-to-needs category by hourly earnings Less than 90 percent of minimum 90 to 110 percent of minimum 110 to 200 percent of minimum More than 200 percent of minimum

0.57 0.20 0.17 0.06

0.52 0.16 0.26 0.06

0.41 0.18 0.32 0.08

0.34 0.17 0.40 0.10

0.25 0.14 0.45 0.16

C. Distributions of workers by family income-to-needs Less than 90 percent of minimum 90 to 110 percent of minimum 110 to 200 percent of minimum More than 200 percent of minimum

0.13 0.08 0.03 0.01

0.15 0.14 0.08 0.01

0.12 0.15 0.12 0.03

0.11 0.11 0.14 0.05

0.18 0.19 0.23 0.16

2,979

5,980

8,852

10,741

24,420

N

Source: Neumark, Schweitzer, and Wascher 2005. Notes: Income-to-needs categories and income measures are reported for year one for each family. Hourly earnings are calculated using annual wage and salary income / {(weeks worked last year) ⋅ (usual hours worked last year)}; this way the full March files, rather than only the ORG files, are utilized. In the first and second panels the columns sum to 1; in the third panel the rows sum to 1 but entries are not shown for income-to-needs greater than 3. The second panel is restricted to families with at least two earners.

nonparametric density estimates that adjust for state and year effects, and yield similar conclusions. Table 2.2 illustrates more clearly how families with incomes initially above the poverty or near-poverty line might be affected by an increase in the minimum wage. Although minimum-wage workers (those earning less than 1.1 times the minimum) account for a very small share of primary earners in families above 1.5 times the poverty line (panel A), it is not unusual for the lowest-paid worker in higher-income families to be paid at or below the minimum wage (panel B). And as shown in panel C, which presents the distribution of workers in each wage category across income-to-needs categories, there is nearly as large a proportion of minimum-wage workers (including those below the minimum) in families with /

35

Making the Work-Based Safety Net Work Better

incomes between 1.5 and 3 times the poverty line as in families between 0 and 1.5 times the poverty line, and actually a greater proportion of minimum-wage workers in families with incomes-to-needs between 1.5 and 3 than below the poverty line. Thus, the evidence that minimum-wage increases cause somewhat higher-income families to fall below the near-poverty line could easily reflect job losses among low-wage workers in these families, and calculations reported in Neumark, Schweitzer, and Wascher (2005) indicate that the numbers of such secondary workers make the estimated effects just discussed quite plausible. At the same time it is worth emphasizing that the research literature does not yet include evidence that directly estimates the effects of minimum wages on workers classified by both wage levels and family incomes. That is, the evidence just discussed suggests that the adverse effects of minimum wages tend to fall on low-wage workers in low-income families, but that is an inference from the effects of minimum wages on the distribution of family incomes rather than a conclusion from direct analysis of how minimum wages affect particular subpopulations. The failure to find beneficial distributional effects of minimum wages is consistent with the broader literature on this question, which is summarized in table 2.3.18 All of these studies rely on CPS data, but they differ in terms of sample period studied, subgroups considered, the measurement of income (before- versus after-tax), and the measures of the income distribution (poverty rate, squared poverty gap, and income inequality measures). Overall, the evidence can be viewed as leading to one of two conclusions, depending on exactly what specifications and approaches one prefers. Either there is no evidence that the minimum wage has beneficial distributional effects, or the minimum wage harms those at the bottom of the income distribution. In essentially no case, though, is there evidence that minimum wages help poor or low-income families.19

Living Wages In the mid-1990s, political support for minimum wage floors was manifested in a new arena: local governments. In cities and other local jurisdictions across the country, campaigns arose in support of “living wage” proposals, and governments adopted them by the score.20 Table 2.4 provides examples of living-wage laws for eight of the largest cities where living wages have been implemented. Clearly many of these are quite a lot higher than minimum wages in the respective states (column 2), and there are considerably higher (and lower) living wages in other cities. On the other hand, living-wage laws have much narrower coverage than state minimum wage laws. As shown in column 3—and as is true more generally of living-wage laws—nearly all living-wage laws cover city contractors, and about half also cover companies that receive financial assistance from cities (such as subsidies and tax abatements). In contrast, living wages rarely apply to city employees. Coverage estimates are very hard to come by, especially for living-wage laws that cover financial assistance recipients, for which city-level information is typi-

36

/

Update of Card and Krueger analysis, 1988 to 2003

March CPS files, 1990 to 2005

Burkhauser and Sabia (2007)

Sabia (2006)

Data

Employed single mothers aged fifteen to fifty-five

Workers, all families

Workers, all families

Sample

Findings Larger minimum-wage increases associated with poverty reductions, but never significant with controls for overall state employment or unemployment. Slightly stronger evidence of antipoverty effects for sample of workers only (but still often insignificant). Larger minimum-wage increases associated with poverty reductions, but evidence never significant with state unemployment controls. Even for workers, estimated effects near zero and insignificant. No evidence of effects of minimum wages on poverty.

Evidence on the Distributional Effects of Minimum Wages

March 1990 and 1992 CPS files

/

Card and Krueger (1995)

Study

TABLE 2.3

(Table continues on p. 38.)

Burkhauser and Sabia (2007) extend analysis to all single female heads of household, with no significant evidence that minimum wages affect poverty.

Specifications with state employment controls not included, although these entered more strongly in Card and Krueger’s analysis.

Conditioning on employment by studying workers masks potential adverse effects of minimum wages.

Comments

Findings

All families, and sub- Mixed evidence: some estimates groups (femalepoint to minimum wages reducheaded households, ing poverty; but for preferred married couples, specification (after-tax income, white families, using squared poverty gap) black families) estimated effect varies in sign and is never significant. All families For a wide variety of inequality measures (but not all), using after-tax income, minimum wages increase inequality. Evidence is strongest for inequality measures that place more weight on transfers at low end of income distribution. Using pre-tax income, minimum wages are always estimated to increase inequality.

Sample

Source: Author’s compilation based on studies cited in table.

March CPS files, 1981 to 1997

Wu, Perloff, and Golan (2006a)

Data

(Continued)

March CPS files, 1981 to 2000

/

Gunderson and Ziliak (2004)

Study

TABLE 2.3

No year effects included in specifications.

Comments

Alternative Labor-Market Policies TABLE 2.4

New York Los Angeles Chicago Philadelphia

San Diego San Antonio

Detroit San Jose

/

Living-Wage Laws in the Eight Largest Cities, as of 2006

Level

Prevailing Minimum Wage

Coverage

(1)

(2)

(3)

$10 $9.39

$5.15 $6.75

$10

$6.50

Service contractors Service contractors, financial-assistance recipients For-profit contractors in specific categories of workers Contractors; businesses with city leases, franchises, concessions; city employees

150 percent of higher of federal or state minimum wage $10

$5.15

For 70 percent of employees in new jobs: $11.14 for services involving durable goods and $10.86 for services involving nondurable goods. Minimum for all workers is $9.62. $10

$5.15

$12.27

$6.75

$5.15 $6.75

Contractors, financial-assistance recipients Financial-assistance recipients (tax abatements)

Service contractors, financial-assistance recipients Service contractors in specific categories, financial-assistance recipients

Source: Author’s compilation based on data from the Living Wage Resource Center, available at: http://www.livingwagecampaign.org/index.php?id=1958 (accessed November 11, 2006). Notes: In most cases, the required wage level is higher if health insurance benefits are not provided. The living wage if such benefits are provided is reported. The prevailing minimum wage is the higher of the state or federal minimum.

cally decentralized. Estimates of coverage by city contractor provisions are typically below 1 to 2 percent, although there is considerable variation in these estimates, and in some cities coverage is higher because of how the law is specified;21 coverage by financial-assistance provisions of living-wage laws is even less clear. There is ample evidence that living wages raise wages, and also evidence that they cause some employment losses, although not surprisingly there is some controversy about the latter conclusions; see Scott Adams and Neumark (2004 and 2005b). Estimates of wage and employment effects from CPS data are reported in columns 1 and 2 of table 2.5.22

/

39

Making the Work-Based Safety Net Work Better TABLE 2.5

/

Estimated Effects of Living-Wage Laws

Log Wages, Lowest Decile of Wage Distribution (Elasticity) (1)

Employment, Lowest Decile of Predicted Wage Distribution (2)

Probability that Family Income Is Below Poverty Line (3)

All living-wage laws Log living wage, lagged twelve months

0.040

–0.053**

–0.035**

Financial-assistance livingwage laws Log living wage, lagged twelve months

0.067*

–0.076**

–0.024*

Dependent Variable

Contractor-only living-wage laws Log living wage, lagged twelve months

–0.006

–0.027

–0.038

N

46,374

116,466

142,421

Source: Adams and Neumark 2004. Notes: The data on labor-market outcomes and other worker-related characteristics come from the Current Population Survey (CPS) monthly Outgoing Rotation Group files (ORGs), from January 1996 through December 2002, and the CPS Annual Demographic Files (ADFs), from 1996 through 2002, for individuals or families residing in MSA’s, in city-month cells with twenty-five or more observations. The data for the first two columns cover 1996 to 2002, and for the last column cover 1995 to 2001. The regressions include controls for city, year, month, minimum wages, and other individual-level controls in the wage and employment specifications, and controls for city, year, and minimum wages in the poverty specification. All specifications also allow differential linear time trends for cities passing or not passing living-wage laws, or passing different types of laws. The entries in the first row are from a specification with a single livingwage variable, and the entries in the second and third rows are from a specification interacting the livingwage variable with dummy variables for the type of living wage. The coefficients for the log wage equation are from log-log specifications, and hence are elasticities. The coefficients from the employment and poverty regressions measure the change in the share employed or poor in response to a one-unit increase in the log living wage (or a 100 percent increase). Reported standard errors are robust to nonindependence (and heteroscedasticity) within city cells. * p ≤ .10; ** p ≤ .05

What about distributional effects? Results from CPS analyses are reported in column 3 of table 2.5. The evidence yields negative point estimates (implying poverty reductions) for both contractor-only and the broader financial-assistance living-wage laws, but only the estimated effect of financial-assistance living-wage laws is statistically significant (at the 10 percent level). For the latter, the estimated coefficient of –.024 implies that a one log unit (100 percent) increase in the livingwage reduces the poverty rate by 2.4 percent.23 Relative to an 18.6 percent poverty rate, this represents a 12 percent reduction, or an elasticity of –.12. This seems like a large effect, given a wage elasticity for low-wage workers below .1. However, 40

/

Alternative Labor-Market Policies

the anti-poverty effects of living wages likely do not come from lifting families from well below the poverty line to well above it, but instead from nudging families over the poverty line; and these average wage effects are likely manifested as much larger gains concentrated on a possibly quite small number of workers and families.24 Note that these positive distributional effects are different from the adverse (or perhaps simply nonexistent) distributional effects of minimum wages. There is no necessary contradiction in these findings, however. Although economic theory predicts that raising mandated wage floors will lead to employment reductions—and the evidence from both minimum wages and living wages is consistent with this prediction—theory makes no predictions regarding the effects of mandated wage floors on the distribution of family incomes, or on poverty specifically. The distributional effects depend on both the magnitudes of the wage and employment effects (and other effects), and on their incidence throughout the family income distribution. The gains and losses from living wages may be of quite different magnitudes, and fall at different points in the distribution of family income, than do the gains and losses from minimum wages—depending in part on the types of workers who are affected by these alternative types of mandated wage floors. Indeed there is evidence of significant differences in the populations of affected workers. David Fairris et al. (2005) report descriptive statistics for workers directly affected by the living wage, based on their survey in Los Angeles. In this sample, 4 percent are teenagers (their table 3.1). I extracted CPS data for Los Angeles for the same years (2002 to 2003). Overall, among workers the percentage of teenagers is 4.2, very similar to their living-wage sample. However, when I restrict the sample to minimum-wage workers, the share of teenagers is much higher. For example, among those earning between $5.15 and $7.25 (the state minimum wage was $6.75, and the federal minimum $5.15), 14.9 percent are teenagers; focusing only on those earning exactly $6.75, the percentage is 14.1. Thus, these data suggest that workers affected by the living wage, although surely less skilled than the average, are quite similar to the overall workforce in terms of the age distribution, with about 4 percent teenagers. In contrast, for workers likely affected by a minimum wage, the share of teenagers is more than three times as high. This presumably helps to explain why living wages have more beneficial distributional effects than do minimum wages, because in the former case there is less scope for the gains to be concentrated among teenagers (as well as young adults aged twenty to twenty-four, who also represent a high percentage of minimum-wage workers, and together with teens about one-third of minimum-wage workers in Los Angeles). It would clearly be useful, however, to better understand how the different distributional effects arise.

Minimum Wages and Skills The evidence on minimum-wage (and living-wage) effects discussed to this point focuses on short-run effects, typically effects at most a year after minimum-wage increases. There are, however, potential effects of minimum wages in the longer/

41

Making the Work-Based Safety Net Work Better

run, through effects of minimum wages on the acquisition of skills via training, schooling, and work experience. Policymakers should be more concerned with how minimum wages affect long-run earnings than with their effects on young adult and especially teen employment, especially with respect to promoting economic self-sufficiency. Although theoretical predictions are ambiguous, minimum wages may lower training among young workers (see Acemoglu and Pischke 2003; Hashimoto 1982; Feldstein 1973). There is evidence from CPS data that minimum wages reduce formal training for those twenty to twenty-four years old (Neumark and Wascher 2001a). Mixed evidence is reported by Fairris and Roberto Pedace (2004), whereas Daron Acemoglu and Jörn-Steffen Pischke (2003) find no evidence of effects on training one way or the other. Overall, I regard the evidence on the effects of minimum wages on training as pointing to possible adverse effects, but hardly conclusive. Minimum wages may also affect schooling. The impact could be positive or negative (see Cunningham 1981; and Ehrenberg and Marcus 1980), depending on how minimum wages affect the returns to searching for employment versus the returns to further education.25 For example, if the minimum wage leads to a relative increase in the demand for more-skilled labor, the price of more-educated labor should rise, which may encourage some youths to stay in school. But the minimum wage also raises the wage of youths who leave school (if they find a job), so that the net effect of a minimum-wage increase could be to lower the return to an extra year of schooling, at the relevant margin. Recent evidence on the effects of minimum wages on schooling of teenagers in the United States points to negative effects. Card (1992) reports a significant negative difference-in-differences estimate of the California minimum-wage increase in 1988 on the teenage enrollment rate, with or without demographic controls. Neumark and Wascher (1995) estimate a model of minimum-wage effects on employment as well as enrollment, and find that a higher minimum wage leads to little change in the proportion enrolled but not employed, a significant negative effect on the proportion enrolled and employed (elasticity of –.47), a weak positive effect on the proportion not enrolled but employed (elasticity of .14), and a significant positive effect on the proportion idle (elasticity of .64). These estimates imply a negative effect of the minimum wage on the proportion enrolled in school. In updated estimates for a later sample period and using a better enrollment measure, Neumark and Wascher (2003) find robust evidence of negative effects of minimum wages on teenage enrollments.26 Similar results are reported in Duncan D. Chaplin, Mark D. Turner, and Andreas D. Pape (2003), based on data on the entire population of U.S. public schools. To assess the overall longer-run influences of minimum wages via training, schooling, work experience, and so on, Neumark and Olena Nizalova (2007) estimate the effects of exposure of workers to higher minimum wages when they were young and unskilled—and hence when minimum wages were most likely to be binding—on outcomes for these individuals when they are somewhat older (twenty-five to twenty-nine years old). The estimates indicate that adults exposed

42

/

Alternative Labor-Market Policies

to minimum wages as teens or young adults subsequently have lower wages and earnings. The effects are not trivial. For example, teenagers’ exposure to the average higher minimum wage (for states with above-federal minimum wages) is estimated to reduce their adult earnings by .8 to 1.8 percent, and similar exposure for those twenty to twenty-four is estimated to reduce adult earnings by 1.7 to 2.3 percent. This latter evidence comes from only one study, and it remains to be seen whether other studies reach similar conclusions. In the meantime, however, the findings suggest that minimum wages may be particularly unhelpful for boosting economic self-sufficiency in the longer run, presumably by reducing the accumulation of skills that lead to higher earnings as an adult.

SUBSIDIZING EMPLOYMENT The minimum wage is often contrasted with the Earned Income Tax Credit (EITC) in discussions of policies to increase economic self-sufficiency. The comparison is natural, because the EITC subsidizes earnings for low-income working families and creates incentives for employment among families with no workers, and thus has much the same goals as suggested by the rhetoric, if not the reality, of minimum wages. (Of course it also affects income above and beyond earnings via the direct payments to families made under the program.) An alternative approach is to subsidize employers for hiring from within particular groups of workers. I discuss these in turn.

The EITC EITC payments are determined by four parameters. The earnings credit establishes a subsidy rate for earnings. Currently, the federal credit rate for a family with two or more qualifying children is 40 percent. There is a maximum benefit level, which for the same type of family was $4,536 (in 2006—this is indexed). There is a “plateau,” or an income range over which the maximum benefit remains fixed (in 2006, from $11,340 to $14,810). And finally, there is a phase-out rate at which the credit is reduced as income rises (currently 21.05 percent).27 The Earned Income Tax Credit has expanded considerably, along two dimensions. At the federal level, the credit rate increased sharply over the 1990s, rising from a rate of 14 percent (with two children) in 1990 to 40 percent in 1996, where it has remained since. In addition, a number of states introduced their own EITC programs; these typically specify a percentage supplement to the federal EITC that is provided to families by the state. The number of states with an EITC rose from seven in 1996 to nineteen (plus the District of Columbia) in 2007.28 The EITC generates a variety of incentives regarding labor supply and other behavior. Good surveys are provided by Saul D. Hoffman and Laurence S. Seidman (2003) and V. Joseph Hotz and John Karl Scholz (2003), and the discussion here is cursory. Regarding labor supply, for families in the phase-in range (that is, earn-

/

43

Making the Work-Based Safety Net Work Better

ing less than the maximum credit), theory predicts that the EITC has an unambiguous positive influence on employment, because the EITC raises the effective wage, and for those previously non-employed there is only a substitution effect and no income effect.29 Hours effects are more ambiguous. On the phase-in range, there are offsetting income and substitution effects. On the plateau, there is only a negative income effect. And on the phase-out range, both income and substitution effects create incentives for reduced hours. The research is fairly unambiguous in indicating that the EITC boosts employment of single mothers.30 Hours effects for those already working but with somewhat higher income (perhaps a working spouse) appear to be modestly negative. The implication of these findings is that the EITC is likely to boost the incomes of low-income families. Moreover, as the preceding discussion makes clear, the EITC targets low-income families; this contrasts sharply with the minimum wage, which of course targets low-wage individuals. Given the weak link between the two, we might expect that the EITC is more effective than the minimum wage at reducing poverty and helping low-income families. At the same time, it is important to note that the EITC does not target poor families perfectly. Most important, the break-even point (at which benefits have fallen to zero) occurs well above the poverty line. Part of the reason for this is that if benefits were phased out too quickly, stronger labor supply disincentive effects (in terms of hours) would be created. In addition, because the EITC is based on income rather than wages, it is possible that it sometimes subsidizes workers who have high skills but work low hours.31 The EITC will fail to reach families with such low incomes that they do not file income tax returns, and they may sometimes subsidize higher-income families in which the adults are unmarried but cohabiting, whose joint income if they married would make them ineligible. Nonetheless, calculations suggest that the EITC targets reasonably well—with very few dollars going to families earning in excess of twice the poverty line (Liebman 1998) and nearly one-half of payments going to poor families (Scholz 1994)—although these calculations are based on simulated policy effects rather than before-and-after analyses. Neumark and Wascher (2001b) use CPS data to estimate how changes in the EITC affect transitions into and out of poverty (and among other income-to-needs categories); the study exploits state policy variation, captured in the credit rate in the phase-in range, which varies by year and state and with the number of children. It would be unsurprising if the EITC lifts families out of poverty when both earnings and EITC payments are considered. However, the study foregoes using estimated EITC payments (and other transfers), and instead just studies earned income; it therefore asks, for example, whether increased generosity of the EITC raises the probability that a family earns its way out of poverty. Such evidence would suggest that the incentives created (and not only the checks written) because of the EITC are pro-work and antipoverty. Indeed, the evidence suggests that increases in EITC generosity raise the probability that family earnings rise to above the poverty level, and also raise the earnings, on average, of families with children that are initially below the poverty line.32 In contrast, there is no effect of pushing near-poor families into poverty (based on earnings), stemming from adverse effects on hours or on

44

/

Alternative Labor-Market Policies

earnings of those initially near-poor. The estimates imply that the average increase in the federal credit rate over the sample period (4 percentage points) reduces the poverty rate by about .029, or nearly 3 percentage points.33 Given the conclusions regarding minimum wages summarized earlier, it is fairly clear that the EITC has better beneficial distributional effects than the minimum wage, and direct comparisons in Neumark and Wascher (2001b) affirm this conclusion. A study by Ximing Wu, Jeffrey M. Perloff, and Amos Golan (2006a) also compares the distributional effects of minimum wages and the EITC, as well as other policies, including income tax rates, unemployment insurance, Supplemental Security Income, Aid to Families with Dependent Children (AFDC), Disability Insurance, and Food Stamps. Their evidence points quite strongly to the conclusion that the maximum EITC benefit reduces inequality. Interestingly, though, they find statistically significant negative effects of the EITC benefit on inequality for all specifications except variants of the Atkinson index (Atkinson 1970) that place relatively more weight on the low end of the income distribution, which the authors suggest occurs because most of the benefits of the EITC actually accrue to families on the plateau and in the phase-out range. And for those families with no workers for whom the EITC does not induce employment, the EITC of course delivers no benefits, which is a potential shortcoming of the policy.34 Finally, a few other issues merit comment. First, despite the evidence of beneficial distributional effects of the EITC, it may not be effective for the very bottom of the family income distribution, although that is almost surely true of the minimum wage as well. That is, both of these policies aim to raise incomes of those who work—although one encourages employment and one may have the unintended effect of reducing it. But other policies—more likely those focused on disability—are needed to deal with the income needs of families with no workers or workers facing work limitations. Second, as pointed out by Michael W. Horrigan and Ronald B. Mincy (1993), among others, the EITC offers virtually no benefits to unrelated individuals (the maximum credit in 2007 was $428, for those aged twenty-five to sixty-four). From a distributional perspective, we may be particularly concerned with increasing economic self-sufficiency among families with children. However, recent work has suggested reasons why we may want to consider expanding the EITC to single unrelated individuals, even if this policy seems counterintuitive (Berlin 2007) from the perspective of past antipoverty efforts. These arguments are considered in more detail in the concluding section of this chapter, which considers more fully alternatives to the policies already in place. Third, I have posed the discussion so far in terms of comparing the minimum wage to the EITC as alternative policies to increase earnings. There are arguments suggesting that this “either-or” comparison is inappropriate, and that instead the minimum wage and the EITC may be complementary, with one making the other more effective. One such argument is that a higher minimum wage may reduce the distortionary impact of the EITC on labor supply. In particular, a higher minimum wage enables a family to achieve the same level of income (earnings plus

/

45

Making the Work-Based Safety Net Work Better

EITC) at the maximum EITC credit with a smaller EITC payment. This in turn allows a lower marginal tax rate over the phase-out range of the credit, which could reduce the associated labor supply disincentives (Blank and Schmidt 2001). This argument and others about interactions between the minimum wage and the EITC are discussed in detail in Neumark and Wascher (2009). Some of the evidence quite indirectly suggests that a higher minimum wage might under some conditions enhance the positive distributional effects of the EITC. This is an intriguing possibility. However, the evidence is preliminary, and more research is needed to try to better establish the effects of minimum wage-EITC interactions. But it is worth emphasizing that this is one potential line of argument that could, depending on the results, lead to a more positive assessment of the distributional effects of minimum wages under some conditions.

Wage Subsidies to Employers The EITC subsidizes employment by adding to workers’ or families’ income an amount based on employment, hours worked, and the wage. Over the phase-in range, and with respect to the employment decision, this operates as a simple employee-based wage subsidy. An alternative policy, with which the United States has more limited experience, is employer-based wage subsidies—that is, making payments to employers for employing particular groups of workers. Although there is not a great deal of research on wage subsidies, evidence on their effectiveness has been reviewed by Lawrence F. Katz (1998), who also presents some new evidence. And the two alternative approaches to subsidizing employment have been contrasted by Stacy Dickert-Conlin and Douglas Holtz-Eakin (2000). A couple of key issues arise with using employer-based wage subsidies. First, the target group has to be identified. In principle, the targeting can be narrowly defined to improve upon that generated by the EITC, although of course simply targeting low-wage workers would run into the same distributional problems as with the minimum wage.35 In the United States, wage subsidies have alternatively targeted the hiring of young disadvantaged workers (Job Opportunities in the Business Sector, or JOBS), AFDC recipients (Work Incentives Tax Credit, or WINTC), and low-wage workers (New Jobs Tax Credit, or NJTC).36 The Targeted Jobs Tax Credit (TJTC) targeted these groups and others and was in effect from 1979 through 1994, when it was replaced by the WOTC, which similarly targets multiple groups.37 Under the Job Training and Partnership Act (JTPA), there were temporary wage subsidies to firms providing long-term employment for recipients of on-the-job training. This was the key wage subsidy for economically disadvantaged adults who were not on welfare. JTPA was replaced by the Workforce Investment Act (WIA) in 2000, which does not include wage subsidies. Both reviews of wage subsidies concur that the evidence suggests that narrow targeting is problematic, stigmatizing the intended beneficiaries and consequently making employers less likely to hire them (or at least dissipating the effects of the subsidies). For example, Katz discusses an experimental program in

46

/

Alternative Labor-Market Policies

Dayton, Ohio, under the TJTC, for welfare recipients. The two groups that received vouchers indicating their eligibility to present to employers were less likely to find employment than a third group of job seekers who did not receive vouchers, but technically had the same eligibility. The second issue is program design. In particular, employer-based wage subsidy programs can create strong unintended incentive effects. If the subsidy seeks to have the most impact on generating new employment, then in some way it has to identify and reward either net (positive) changes in employment that would not have otherwise occurred, or new hires directly. The latter approach can create incentives for churning employees.38 The former approach may not create incentives for churning, but it imposes the need for a good deal of information from firms (to try to determine, on the basis of some rule, what employment would have been otherwise). In addition, this approach can generate incentives for employment variation (since increases are subsidized but reductions are not penalized). A simpler scheme can simply subsidize all employment (perhaps in targeted groups), but this would reward employers in large part for what they are doing already (“windfalls”), and hence entail much greater expense and likely have less impact. In general, any employer-based subsidy is likely to be rather complicated, in terms of both identifying eligible workers and determining the subsidy paid to firms. As a consequence, subsidies entail serious administrative constraints on firms. These constraints likely explain the very low utilization of employer-based wage subsidies when they have been implemented. How do wage subsidies compare with the EITC, which effectively pays workers to subsidize their employment (and hours)? In the simplest textbook world, the wage and employment effects of a wage subsidy do not depend on whether the subsidy is paid to employees or employers. However, as Dickert-Conlin and Holtz-Eakin (2000) emphasize, things are more complicated. In addition to the fact that the EITC is not a simple wage subsidy, they emphasize factors that likely diminish the effectiveness of wage subsidies in particular. One issue is that of stigma, already discussed, which implies that any wage subsidy is less effective than the “posted” subsidy because of the negative information it may convey about potential employees. The EITC does not have stigma effects, since the employer typically has no idea whether an employee is eligible for or receiving the EITC, and it has good but, as noted, not ideal targeting. 39 Wage subsidies that do not target specific groups but instead subsidize only an initial amount of wages reduce stigma, although of course the efficacy of targeting is then reduced. A second issue is administrative and compliance costs. As already noted, these can be substantial for employer subsidies. The EITC, on the other hand, is easily administered through the tax code, although there is a sizable potential cost from fraudulent claims associated with claiming children on tax returns when they did not in fact reside with the filer for the half of the year required by the law. The simplest model also breaks down when we consider incentives on the extensive (employment) versus intensive (hours) margins. A program that does more on the extensive margin likely delivers more of the intended impact, but Dickert-Conlin and Holtz-Eakin (2000) suggest that employer subsidies may have

/

47

Making the Work-Based Safety Net Work Better

more impact on the intensive margin, for which they suggest that supply is less elastic and demand more elastic, in which case more of the benefits of the subsidy go to employers. Despite problems with employer-based wage subsidies, Katz (1998) presents some evidence indicating that the TJTC did boost employment of disadvantaged youths, and discusses evidence indicating positive and persistent program impacts from JTPA when the training was combined with job search assistance, especially for adult female welfare recipients. This leads him to conclude that wage subsidies combined with training and job development assistance can help disadvantaged adults, but, on the basis of the evidence on stigma and low utilization, to express more skepticism (while still suggesting modest benefits) of other narrowly targeted, stand-alone programs. In more recent evidence on the WOTC, Sarah Hamersma (2005b) concludes that any employment effects are small, if they exist at all, and are hard to establish using the existing evidence. She does find positive effects on earnings (of around 10 percent), although only on the job paying the credit, and not over the course of the year after a worker starts a subsidized job. Dickert-Conlin and Holtz-Eakin (2000) favor the EITC over wage subsidies, citing evidence on the positive labor supply and poverty-reducing effects of the EITC, as well as the limited evidence of effectiveness and utilization of employer-based subsidies. I would suggest that we do not yet have sufficient evidence on the impact (and costs) of wage subsidy programs to be able to make definitive statements, and both reviews are similarly cautious. However, the considerations discussed here suggest that there is little compelling basis for preferring employer- over employee-based subsidies, unless we can not only identify effective employer-based wage subsidy programs but also demonstrate that for particular groups they may be more effective than the EITC, or enhance the positive benefits of the EITC. Note that some of the evidence Katz discusses suggests possible cases like these, in particular for marginalized groups such as disadvantaged adults in need of training, or welfare recipients. (This echoes some of the concerns raised earlier regarding the inability of the EITC to provide much benefit at the bottom end, as it delivers no benefits to the nonemployed.) His suggestion that employer-based wage subsidies appear most promising when combined with other job-related services may point to the need for particular interventions to provide training and job skills to those for whom simply strengthening financial incentives to work may not be enough.

INCREASING SKILLS VIA SCHOOL-TO-WORK I now turn to discuss public policy that seeks to improve earnings and hence economic self-sufficiency through increasing skills. As noted earlier, my discussion focuses on school-to-work policies. Because school-to-work policies are, by intention, effective mainly for young people, an important element omitted from this discussion is efforts to increase skills among adults already in the labor market.

48

/

Alternative Labor-Market Policies

As already noted, there is an extensive literature on training programs, some targeted to adults. Although I do not discuss the training literature in any detail, it is worth pointing out that evidence compiled in that literature tends to find that traditional job-training programs have some beneficial effects in terms of increasing earnings for disadvantaged adults, but quite limited effectiveness for less-advantaged youths.40 This may provide an additional motivation for focusing on schoolto-work programs, as an alternative and possibly more effective means of intervening with teenagers to increase skills and thus earnings.

School-to-Work Background and Policies School-to-work policies were encouraged to reduce the “churning” or “milling about” experienced by some youths in the United States upon their entry into the labor market, reflected in initial periods of joblessness or a series of dead-end jobs (U.S. General Accounting Office 1990). Researchers and others advocated that the United States adopt a more orderly school-to-work system, like that of the German apprenticeship system or the informal contracts between Japanese schools and employers (see, for example, Commission on the Skills of the American Workforce 1990; Hamilton 1990). This policy debate provided much of the impetus for the 1994 School-to-Work Opportunities Act (STWOA), which provided $1.5 billion in federal funding to support the creation of an integrated system of youth education, job training, and labor-market information, to provide a faster and more successful transition from school to stable employment in higher-paying jobs.41 In addition, school-to-work practitioners and advocates commonly argue that school-to-work programs like those encouraged by the STWOA are especially helpful for the “forgotten half”—the non-college-bound among whom the less-advantaged are concentrated. However, in studying school-to-work programs it is important not to focus solely on the STWOA. As described in more detail in Neumark (2007, chapter 1), school-to-work policies culminating in the STWOA developed from a rather long history of policies addressing the schoolto-work transition. What distinguishes the STWOA is its efforts to integrate academic and vocational education for the traditional target audience of those bound for at most two-year degrees, as well as those bound for four-year colleges and universities (Cohen and Besharov 2002).42 A second effort toward integrating academic and vocational skills was the development of “career academies,” which are “schools within schools” that integrate academics with general job readiness and preparation in a particular career area. The STWOA was not reauthorized after its initial five years, and the federal effort was essentially abandoned by the Bush administration in favor of No Child Left Behind (NCLB), which emphasizes measurable academic outcomes and assigns a central role to standardized testing in grades K through twelve. Although educational reforms focusing on school quality can be viewed as seeking to enhance labor-market success via better educational preparation, there is some basis for asking whether test-based reforms do as much to prepare students for careers

/

49

Making the Work-Based Safety Net Work Better

as do school-to-work efforts, especially for the forgotten half. In this section of the chapter I examine the effectiveness of school-to-work efforts in improving prospects for economic self-sufficiency. A comparison of school-to-work efforts and test-based educational reforms is more challenging and is not taken up here.

The Effects of School-to-Work In this section of the chapter, I provide a broad overview of earlier research on the effects of school-to-work, followed by a more detailed discussion of more recent research studying the effectiveness of school-to-work in general and in specific contexts—such as career academies. EARLIER RESEARCH David Stern et al. (1995) provide a thorough compendium of earlier research on school-to-work programs. The research summarized in that volume offers little persuasive evidence of positive impacts of these programs on adult labor-market outcomes. First, few studies have focused on labor-market outcomes more than a year or two after completion of the programs, and those that do provide limited evidence that over a period of a few years beneficial effects of some types of school-to-work programs dissipate, as comparison group members find good jobs on their own. Second, many of these studies do not construct a reasonable comparison group, let alone consider the problem of selection into programs on the basis of unobserved characteristics that might also be correlated with outcomes. Third, even studies that attempt to construct a good comparison group find no beneficial short-term labor-market effects, with the possible exception of students who remained with the employer with whom they “apprenticed” during the program. Finally, some of the evidence suggests that school-to-work programs may discourage postsecondary education. A major report on the STWOA by Mathematica Policy Research (Hershey, Silverberg, and Haimson 1999) did little to advance our understanding of program impacts; in fact, the main goal of this report was not to provide a program evaluation. Nonetheless, the report does present some evidence that is intended to speak to the effects of school-to-work programs. For example, the report notes that students in paid positions arranged as part of school-to-work programs are employed in a wider array of industries and receive more training than other students in paid positions, and concludes that “schools develop positions in a wide range of industries, increasing the chances that students can work in a setting relevant to their career interests” (89). However, nothing in the evidence implies that students who found these jobs as part of school-to-work programs would not have found the same types of jobs absent such programs; students most likely to do so may simply have sorted into school-to-work programs. Students do, however, report that school-to-work activities helped them sharpen their career goals; whether this translates into concrete gains was not examined. Paralleling this view of the earlier evidence, a subsequent survey of published academic research on school-to-work across the United States generally supported the claim that lit-

50

/

Alternative Labor-Market Policies

tle progress had been made in estimating the causal effects of school-to-work programs (Hughes, Bailey, and Mechur 2001). EVIDENCE FROM THE 1997 NATIONAL LONGITUDINAL SURVEY OF YOUTH (NLSY97) The NLSY97 offers researchers opportunities to improve significantly upon the existing research. The NLSY97 covers respondents who were in high school during the period in which the STWOA was in effect, and surveys individuals about “programs schools offer to help students prepare for the world of work.” The school-to-work programs covered in the NLSY97 include job shadowing; mentoring (matching students to an individual in an occupation); cooperative education (combining academic and vocational studies); work in a schoolsponsored enterprise; Tech Prep; and internships or apprenticeships. However, the NLSY97 has not until recently permitted much follow-up as the respondents age into adulthood. In two papers Donna Rothstein and I (Neumark and Rothstein 2006, 2007) explore the effectiveness of school-to-work programs using the NSLY97. The analysis in the first paper is based on data from the first four rounds of the NLSY97. When the first round was administered, in 1997, respondents were age twelve to seventeen. With the second round we could begin to observe some respondents who have left high school, but we get many more such observations with the third and fourth rounds, and therefore focus on educational and employment outcomes measured as of the third or fourth rounds (and in the 2007 paper, the fifth round). The core empirical framework is estimation, at the individual level, of the relationship between employment or enrollment in the post-high-school period, and participation in school-to-work during high school. The analysis is based on dichotomous choice models for employment and enrollment; both of these activities are considered important to later labor-market success, in particular relative to the alternative of being neither in school nor employed. We estimate models for employment and enrollment outcomes as determined by participation in a number of different school-to-work programs, as well as a vector of individual, family, or school-level controls. However, this research has to confront the potential for endogenous selection into school-to-work participation. For example, individuals with the highest expectations of work after high school may choose to participate in these programs, perhaps because they are more interested in learning about the job market. When actual work behavior is observed, then, we would find that those who participated in school-to-work have higher employment rates, but the positive relationship between school-to-work participation and employment is at least partly noncausal. Alternatively, schools with student populations that tend not to go to college may be more likely to offer school-to-work programs and to have students with lower college enrollment, in which case, again, the association between school-to-work participation and later outcomes might not be causal. A common approach to the endogenous selection problem with longitudinal data is to implement a first-difference estimator based on observations before and after program

/

51

Making the Work-Based Safety Net Work Better

participation. But because we are studying program effects on individuals’ first labor-market experiences, or on further school enrollment of those already enrolled, there are no meaningful observations on the outcomes of interest prior to participation, and hence such estimators are inapplicable. Instead, we use an extensive set of controls for the factors that might affect the dependent variables and also might be correlated with school-to-work participation. In addition to standard demographic information, the NLSY97 includes data on living arrangements and family structure and income, test scores (from the Armed Services Vocational Aptitude Battery, or ASVAB), and self-reported measures of school behavior (whether the respondent was threatened at school, or had gotten into a physical fight at school, and information on absences or tardiness at school). Even more useful are the respondent’s self-reported subjective probabilities for future education and employment, including receipt of a high school diploma by age twenty; obtaining a four-year-college degree by age thirty; and working over twenty hours per week at age thirty. Including these expectations variables can be viewed as mimicking the longitudinal estimator that we would like to have, as these controls capture some measures of an individual’s propensity for later enrollment or employment prior to participation in school-to-work.43 We also found that school offerings of school-to-work programs were correlated with characteristics of schools and their student bodies, such as truancy, pregnancy, and alcohol and drug use. These associations suggest that if we use individual-level school-to-work participation to identify the effects of school-towork, but the variation in participation stems partly from variation across schools, then we may obtain biased estimates. The NLSY97 has one additional feature that we can exploit to address this problem. Specifically, there are data on multiple students in the same school, allowing school fixed effects to be added to the equations to control for unobserved factors that are common to students within a school. Of course, since individuals within schools differ from one another, in the within-school estimation attention must still be paid to individuallevel heterogeneity. Key results are presented in table 2.6. As it turns out, the estimates are quite similar across the alternative specifications—including just the basic demographic controls, the more detailed ones listed in the table, the proxy variables measuring respondents’ work and schooling expectations during high school, and the school fixed effects. Consequently, I simply summarize the overall results. Looking first at college education, the estimates reveal considerable heterogeneity in the effects of different types of school-to-work programs. There is some evidence that job shadowing and mentoring programs are associated with a significantly higher likelihood of some college education, and robust evidence of a positive effect of school enterprise programs.44 The estimates are often quite sizable. For example, in panel A, with the basic controls, the estimated differentials for mentoring and school enterprise programs range from about .07 to .11; relative to the sample proportion of .50 with some college, these estimates imply increases of 14 to 22 percent in the probability of college attendance. On the other hand, Tech Prep programs are associated with a significantly lower likelihood of college education.

52

/

Alternative Labor-Market Policies

The employment results are more robust across specifications and samples, generally pointing to statistically significant positive effects of coop (cooperative education) programs, with effects near .06 to .10, and also evidence of positive effects of internship or apprenticeship programs. Relative to a sample proportion currently employed of .63, the estimates imply that coop programs are associated with approximately 10 to 16 percent increases in the probability of post–high school employment, and internship or apprenticeship programs with increases of about 8 to 18 percent.45 A key finding is that the school-to-work programs that boost college attendance (job shadowing, mentoring, and school enterprise) do not adversely affect employment, and conversely, that programs that boost employment (coops and internships or apprenticeships) do not reduce college attendance. This suggests that, on net, these school-to-work programs increase skill formation; if, in contrast, they tended to reduce employment when enrollment rises, or vice versa, the conclusions would be more ambiguous. The one piece of evidence on school-towork programs that appears in a more negative light is the estimated negative impact of Tech Prep on college enrollment. Paralleling the earlier discussion, if there is any trade-off with employment so that work increases, this negative enrollment effect might be viewed less negatively. It turns out that there is a positive, although only weakly significant, effect of Tech Prep on the likelihood of full-time work conditional on employment, which is roughly the same size (but opposite sign) as the estimated effect on enrollment. This suggests that the negative effect of Tech Prep on schooling is roughly offset by a higher incidence of full-time work. Nonetheless, as the returns to schooling in the form of higher wages typically outweigh the returns to experience—and even more so with regard to the return to full-time versus part-time experience—it is difficult not to view the negative effect of Tech Prep on schooling in a somewhat adverse light. Finally, it bears repeating the reservation that these estimates are only suggestive of longer-term beneficial effects; more definitive, longer-term analyses await more data. SCHOOL-TO-WORK PROGRAMS AND THE “FORGOTTEN HALF” Neumark and Rothstein (2007) use the NLSY97 data to explore the differential effects of school-to-work program participation on disadvantaged and minority youths. The analysis proceeds in two steps. First, to operationalize the “forgotten half,” we estimate a reduced form model for attending college. We do this without incorporating information on school-to-work participation, to establish the ex ante probabilities of college attendance (on the basis of which schools or policymakers might target school-to-work efforts). We use the estimates of this model to distinguish between those in the top and bottom halves of the distribution of the predicted probability of college attendance, interpreting the latter as the “forgotten half.”46 We then estimate regression models for the effects of participation in various school-to-work programs on a number of postsecondary education- and employment-related outcomes, allowing for separate effects of program participation for those in the top and bottom halves of the predicted probability of college attendance—in other words, separate effects for the forgotten half. These esti-

/

53

Making the Work-Based Safety Net Work Better TABLE 2.6

/ Linear Probability Estimates of Effects of School-to-Work Participation on College Attendance and Employment

Some College (1) A. Detailed control variables Job shadowing Mentoring Coop School enterprise Tech prep Internship or apprenticeship Includes demographic controls Includes controls for living arrangements and family structure, ASVAB, and school behaviors B. Expectations proxies Job shadowing Mentoring Coop School enterprise Tech prep Internship or apprenticeship High school diploma by age twenty Four-year degree by age thirty Work over twenty hours per week at age thirty

54

/

.015 (.026) .066* (.036) –.019 (.028) .112*** (.037) –.059** (.030) .045 (.032) Yes

(2) .037 (.023) .026 (.031) .007 (.026) .088*** (.033) –.042 (.030) .021 (.030) Yes

Employment (1’) .006 (.025) –.035 (.033) .079*** (.028) .025 (.037) –.000 (.028) .053* (.030) Yes

Yes .024 (.028) .019 (.039) .021 (.031) .113*** (.040) –.046 (.038) .012 (.036)

.014 (.027) –.008 (.038) .030 (.030) .104*** (.039) –.016 (.035) .016 (.035) .010 (.072) .428*** (.035) .054 (.064)

(2’) –.000 (.025) –.029 (.033) .078*** (.028) .016 (.037) –.007 (.028) .059* (.030) Yes

Yes .017 (.030) –.007 (.041) .055* (.033) –.025 (.048) .031 (.033) .052 (.037)

.018 (.030) .008 (.041) .052 (.033) –.019 (.049) .030 (.033) .052 (.037) .025 (.090) –.101** (.043) .226*** (.079)

Alternative Labor-Market Policies TABLE 2.6

/ (Continued)

Some College (1) C. School fixed effects Job shadowing Mentoring Coop School enterprise Tech prep Internship or apprenticeship Hausman test for excluding school fixed effects, p-value School fixed effects included

.035 (.027) .018 (.034) .004 (.031) .091** (.038) –.070** (.036) .038 (.036)

(2) .063** (.030) .048 (.039) –.013 (.035) .133*** (.048) –.095** (.040) .055 (.041) .18 Yes

Employment (1’) –.019 (.028) –.031 (.037) .075** (.033) –.002 (.047) .011 (.032) .116*** (.035)

(2’) –.026 (.035) –.057 (.047) .102*** (.037) –.018 (.056) .036 (.041) .073* (.043) .24 Yes

Source: Author’s compilation based on Neumark and Rothstein 2006. Notes: School and work outcomes are measured as of the post–high school interview (1999 or 2000). The standard errors allow for general heteroscedasticity and nonindependence within schools. The sets of control variables are detailed in Neumark and Rothstein (2006). All of the specifications in panels B and C include the demographic, living arrangement or family structure, ASVAB, and school behavior variables that are included in columns 2 and 2’ in panel A. * p < .10; ** p < .05; *** p < .01

mates are then used to test which types of school-to-work programs are particularly effective at boosting postsecondary outcomes for the forgotten half. We use the same data as in the earlier paper, but extended through the fifth round. In addition, we explore a richer set of education- and work-related outcomes. Because the findings from Neumark and Rothstein (2006) suggested little evidence of endogenous selection into school-to-work programs in a manner that biases the estimates of program effects, for this analysis the regressions with the detailed controls, but without either the subjective expectations data (available for only a subsample) or school fixed effects, were used. Table 2.7 provides a summary of the results. The table displays the estimated signs of effects for those in the forgotten half, showing all cases for which the estimates are significantly different from zero only for the forgotten half (or significant for both, but with the opposite sign for the forgotten half). The estimates are also broken up into those indicating that school-to-work participation increases skills, and the opposite case. Thus, the entries in the “skill increasing” panels

/

55

Making the Work-Based Safety Net Work Better TABLE 2.7

/ Summary of Effects of School-to-Work Participation on the “Forgotten Half”

Schooling-Related (1) Females Skill increasing Job shadowing Mentoring Coop School enterprise Internship or apprenticeship

Skill decreasing Tech prep Internship or apprenticeship Males Skill increasing Job shadowing Mentoring

Coop

Work-Related (2)

Idle:– – Hours: +

Attended two-year college: ++

Earnings, uncond.: +++ Wage, uncond.: ++ Earnings, cond.: ++ Wage, cond.: ++

Any college: – Training: –

Weeks in school: ++ Any college: + Currently enrolled: + Attended four-year college: ++ Any college: ++ Currently enrolled: +++ Attended two-year college: +

School enterprise Tech prep

Weeks in school: +

Internship or apprenticeship

Attended two-year college: ++

Skill decreasing Internship or apprenticeship

Attended four-year college: – –

Earnings, cond.: +

Idle: – – –

Weeks working: + Weeks idle: – – Weeks idle: –

Currently working: +++ Weeks idle: –

Source: Neumark and Rothstein 2007. Notes: The results shown are those for which the estimated effect was statistically significant at the ten percent level or better only for the forgotten half (or significant with the opposite sign for the forgotten half). The sign is as indicated, appearing three, two, or one times to indicate that the estimate for the indicated group is significantly different from zero at the 1, 5, or 10 percent level, respectively. In all cases, effects that increase schooling, work, skills, or earnings are included in the rows labeled “skill increasing,” and vice versa.

56

/

Alternative Labor-Market Policies

highlight the school-to-work programs that appear to increase schooling, employment, and earnings only for those in the forgotten half. Looking first at the results for women, there is little indication of positive effects on schooling-related outcomes for those in the forgotten half. However, the conclusion is quite different for work-related outcomes, with the most striking finding that internship or apprenticeship programs appear to be particularly effective at boosting wages and earnings for the forgotten half. For men, the situation is somewhat different. In general, there is more evidence that school-to-work programs are particularly beneficial for the forgotten half. Moreover, there is evidence of beneficial effects on both schooling-related and work-related outcomes. With regard to schooling, for five of the six programs there is evidence of positive effects on education, and for the work-related outcomes, there is particularly strong evidence that internship or apprenticeship programs boost employment and decrease idleness among men in the forgotten half, with similar results for school enterprise programs. Overall, then, there is evidence that school-to-work programs are particularly advantageous for men in the forgotten half with respect to both schooling and work-related outcomes, but for women only with respect to work-related outcomes. Thus, the combined evidence from the NLSY97 points to some beneficial effects of school-to-work programs, and suggests that, especially for men, participation in some school-to-work programs increases education and employment among the forgotten half. EVIDENCE ON CAREER ACADEMIES “Career academies” have three elements. First, they combine academic courses that meet college entrance requirements and technical classes that relate to a specific theme, which is typically organized around an industry such as finance, travel and tourism, public service, or information technology. Second, academies engage in partnerships with employers in the industry to provide internships and other work-related experiences outside the classroom. Finally, career academies are typically structured so that groups of students take the same classes together in each grade, and stay with the same group of teachers for at least two years.47 Career academies can be thought of as relatively intensive versions of school-to-work programs. Even some recent studies of career academies (Orr et al. 2007; Stern et al. 2007) suffer from two problems—an inability to address endogenous selection into these academies, and a lack of data following students after they leave high school. Nonetheless, these studies provide descriptive information on the content of career academy programs, and some suggestive evidence that career academies associated with fields that generally require a college degree may boost college attendance. However, the ongoing evaluation of career academies by the Manpower Demonstration Research Corporation (MDRC) (Kemple and Snipes 2000; Kemple 2001, 2004) addresses explicitly the problem of endogenous program participation, as well as later labor-market outcomes.48 The strength of this study is that it is based on random assignment of students to career academies, as participants

/

57

Making the Work-Based Safety Net Work Better

were chosen randomly from applicants to the career academies in the study, with participants and nonparticipants followed for four years (thus far) after leaving high school. The most recent report (Kemple 2004) finds beneficial labor-market effects for male participants, but not for female participants. For example, for male participants relative to nonparticipants the probability of any employment and full-time employment is higher in each of the four follow-up years, with differences of 3.6 to 11.8 percentage points (mostly significant). Average weekly hours in each year are significantly higher, by about four hours, the average wage is significantly higher in most years, by about $.80 (although less in year two), and average monthly earnings are higher in most years, by $166 to $268 (significant in all years but year three, and highest in the last year). There was also some evidence that participants were in better jobs; as of the last quarter, among those employed, participants were more likely to have a health plan, sick leave, paid vacation days, and a retirement plan, although only the vacation day differential was significant. Comparing the estimated differences in employment, hours, earnings, and so on, across each of the four follow-up years, there is no evidence of increasing (or decreasing) differentials between the participants and the controls. The earnings differential, for example, is $206 in year one, $202 in year two, $166 in year three, and $268 in year four. The wage, hours, and employment differentials are, similarly, relatively stable. This does raise an interesting question of what the nature of the effect of career academies is. The evidence does not point to greater career progress of participants. Rather, the differentials that emerge in the first year after leaving high school are largely persistent. For women, interestingly, none of these differences emerged. The estimated differences were much smaller and statistically insignificant, although the point estimates tended to favor participants over nonparticipants. James J. Kemple points out that the young women in the sample were more likely to have children than the young men, with no differences between women who participated and the controls. He suggests that the higher incidence of childbearing “may have limited the extent to which the young women were able to capitalize on the career development experiences that they had in the career academies” (2005, 16). Finally, with respect to labor-market outcomes, the results indicated that the beneficial effects were concentrated among those who came from higher-risk groups, defined as either those with a high likelihood of dropping out, on the basis of characteristics measured prior to random assignment, or those who were not highly engaged in school.49 Of course it is possible that the labor-market gains for men could come at the expense of schooling, which might deliver more long-run returns. The evidence for men suggests many negative schooling-related differentials for participants relative to nonparticipants, although only one is significant—for ever enrolling in postsecondary education (a 6.8 percentage point differential). Other shortfalls for male participants, although none are significant, include completion of a postsecondary credential, receipt of a bachelor’s degree, or receipt of a skills training certificate or license; the estimates range from about 2.3 to 4.2 percentage points.

58

/

Alternative Labor-Market Policies

Kemple argues that the evidence suggests that the “labor-market outcomes for young men did not come at the expense of reducing the prospects for postsecondary education” (2005, 21). I am a bit less convinced of this given the sign patterns of the estimates, although clearly in terms of statistical significance the evidence of positive labor-market effects is stronger. Moreover, as the study reports, for the higher-risk group defined in terms of likely dropout, there was a significant negative effect on postsecondary educational attainment.50 For women there is no evidence of effects on education. Given the potential negative schooling effects, it will be important, if possible, to follow up participants even later, to see whether the positive effects for male participants decline as nonparticipants leave school, although it is highly unlikely that this could erase the reported gains. Finally, one slightly troubling aspect of this study is that the non-academy control group did exceptionally well in terms of high school completion. The rates for the treatment and control groups, respectively, were 81.3 and 83.3 percent, compared with 61.2 percent in a comparable random sample of students in a general curriculum, which Kemple suggests is the group from which academy students typically come. (The study does not attempt to provide similar comparisons for the other outcomes it covers.) In my view, the high achievement of the control group raises questions about the generalizability of this study. In addition, these comparisons put the estimated effects of career academies in perspective, suggesting that even though there were significant gains relative to the control group, the gains may have been relatively small compared to the low achievement and performance of the general population that career academies are intended to help. Nonetheless, this study does provide evidence of positive gains from career academy participation. The evidence from the MDRC study (Kemple and Snipes 2000; Kemple 2001, 2004) clearly represents a major addition to the evidence of beneficial effects of career academies. It also to some extent parallels the findings from the NLSY97 for school-to-work more generally, indicating that school-towork programs deliver labor-market gains for disadvantaged men. SCHOOL-TO-WORK AT POSTSECONDARY INSTITUTIONS All of the research on school-to-work programs discussed to this point, and most of the existing research literature, focuses on high schools. This reflects the policy emphasis. Ann E. Person and James E. Rosenbaum (2007) provide a provocative perspective that suggests that this policy emphasis should be at least partially redirected. In particular, they begin by arguing that linkages between high schools and labor markets are relatively weak, with employers having little regard for high school achievement, and high school staff having little trust of or interaction with employers. Of course school-to-work programs seek to change this. Nonetheless, echoing W. Norton Grubb (2001), among others, Person and Rosenbaum emphasize the potentially important role played by community colleges, and occupational colleges in particular, in the workplace preparation of lower-skilled adults. Person and Rosenbaum point out that nearly half of the students who enter postsecondary education attend community colleges. This fact, coupled with the occu-

/

59

Making the Work-Based Safety Net Work Better

pational focus of many community colleges—based in part on the ability of these local institutions to collaborate with local employers and government—suggests that we should look to community colleges as playing a potentially important role in the school-to-work transition. Person and Rosenbaum’s research seeks to explore the role of two-year postsecondary institutions in creating linkages between school and work. Much of their focus is on differences between private occupational colleges and public community colleges, which is potentially informative about what the latter institutions might do better. More generally, this work can be viewed to some extent as testing the idea that these institutions are a potentially fertile ground for policy efforts focused on the school-to-work transition. The authors provide a mixed qualitative and quantitative study of labor-market linkages among faculty at two-year colleges, comparing seven public community colleges and seven private occupational colleges; the latter are hardly representative, and may actually provide valuable “best practices” models. The qualitative evidence from interviews of program chairs at the two kinds of schools points to a number of dimensions along which labor-market linkages are taken more seriously at the private occupational colleges, even if there are formal responsibilities for labor-market linkages at the public community colleges. These include greater contact and integration between the teaching faculty and career services, greater involvement with advisory committees of local employers, fewer bureaucratic obstacles to changing curriculum to respond to new developments, more individualized and intensive job placement efforts, and a mission more focused on workforce training rather than general education and transfer to fouryear colleges. Analysis of data from a survey of over 4,000 students at these colleges tries to examine quantitatively whether labor-market linkages on the part of teachers or the institutions (as perceived by students) had beneficial effects. Person and Rosenbaum found that labor-market linkages are greater at private colleges and lead to increased effort in school and greater confidence about degree completion. This conclusion is reinforced by evidence from national samples that job placement services at private community colleges are positively associated with degree attainment. Clearly more evidence is needed of the actual effects of school-towork-type interventions at two-year and community colleges. But this work points to the potential importance of focusing policy efforts on two-year colleges in addition to high schools. And one critical question is how two-year and community colleges can better contribute to improving skills among adults already in the labor market.

CONCLUSIONS AND DISCUSSION OF POLICY OPTIONS MOVING FORWARD This chapter does not consider all of the labor-market policies that could potentially improve economic self-sufficiency by increasing earnings. It focuses on

60

/

Alternative Labor-Market Policies

mandated wage floors, the EITC and wage subsidies, and school-to-work programs. It excludes key issues such as welfare reform, training, and education reform, including early childhood investments, without in any way suggesting that these are less important than the topics covered. As noted in the introductory section to this chapter, each of the policies I examine can be considered in the context of a simple supply-and-demand analysis. A minimum wage mandates a movement up the labor demand curve, toward higher wages and lower employment. The EITC pays subsidies to workers to increase employment, and hence shifts the labor supply curve outward, raising employment, potentially lowering the market wage, but increasing the combined wage including the EITC. Wage subsidies instead subsidize hiring by employers, shifting the demand curve out and also raising employment and the wage (part of which is paid in the form of the subsidy). And finally, school-to-work policies aim to increase productivity, also shifting the demand curve out and raising wages and employment. Even in a simple textbook setting, it is unclear which of these policies will do the most to increase income. And once a more realistic view is adopted, with heterogeneous workers who can be affected differently by the alternative policies, predictions about which policies will work become even more complicated, as questions of who gains and loses, and by how much, come to the fore. What are the main conclusions from the empirical analysis of each of these policies? First, the minimum wage is an ineffective policy to promote economic self-sufficiency through higher earnings. It reduces employment of the least-skilled individuals it is trying to help. That in itself does not imply that minimum wages do not on net help. The more telling evidence is that minimum wages do not deliver beneficial distributional effects to poor or near-poor families, and may make them worse off. In addition, minimum wages appear to have deleterious longer-run effects on earnings, presumably through reducing the accumulation of skills. Putting the case succinctly but strongly, it is extraordinarily difficult to discern any case for higher federal or state minimums in order to improve economic selfsufficiency. Do these conclusions imply that I advocate discontinuing the minimum wage? I do not, mainly because the evidence from which the effects of minimum wages are estimated comes from small changes in minimum wages, and hence simply does not support inferences about the effects of large policy changes. By the same token, however, the generally harmful effects I find from small increases in the minimum wage suggest that the target population would be better served by slow erosion of the minimum wage, at least over a modest range, which to a large extent is what happened since about 1980 (until recently), at least at the federal level. Interestingly, though, living wages, which target different workers, present a more favorable trade-off. They still entail disemployment effects but appear to deliver more beneficial distributional effects. Of course the implication of the research on minimum wages is that sharply expanding the coverage of living-wage laws would take us into the territory of minimum wages, with their concomitant

/

61

Making the Work-Based Safety Net Work Better

adverse effects. Moreover, even if the findings on living wages might imply that at much higher levels minimum wages could have some beneficial distributional effects, the costs of raising the wage floor from current levels of minimum wages to the much higher levels typical of living wages could be very high. Labor supply incentives, in particular the EITC, appear to be effective. There seems to be fairly compelling evidence that a more generous EITC boosts employment of single mothers and in so doing raises incomes and earnings of lowincome families. Wage subsidies are the flip side of trying to strengthen employment incentives, but provide the incentives to employers rather than employees. There is some evidence that these subsidies increase employment and earnings. (Certainly if I am convinced that employment for the low-skilled falls in response to a higher minimum wage, I should also think that subsidizing wages boosts employment.) However, as Dickert-Conlin and Holtz-Eakin (2000) emphasize, problems of stigmatization resulting from eligibility for wage subsidy programs can offset some of the gains. Coupling such programs with training and job search assistance may reduce problems associated with stigma and thus increase the benefits of wage subsidies. Another possible means to avoid stigmatization is to pay the subsidy to workers instead of to firms (Scholz 2007)—a policy more like the EITC but based only on low wages.51 Assuming that the low effectiveness of existing subsidy programs is principally due to low participation by firms, but that the effects of a wage subsidy paid to employees would parallel those estimated for the EITC, the employment effect could be sizable (Scholz 2007). Although wage subsidies paid to employees are worth considering, at present there does not seem to be a great deal of political support for expanding them. Moreover, a major effort in this direction entails substantial administrative difficulties. Thus, it may be that increasing the generosity of the EITC for unrelated individuals is a more realistic option for further extending subsidies to employment, especially if the goal is to increase earnings (and incomes) among those not currently eligible for the EITC. One argument for extending the EITC to individuals is to offset some presumed adverse consequences of lower earnings for the less-advantaged—including declining employment rates—stemming in large part from the long-run increase in wage inequality (Berlin 2007). Gordon L. Berlin also points out that there have been declines in marriage and increases in out-of-wedlock childbearing and childrearing, and that these changes may have occurred in part because declining earnings of men made them less attractive marriage partners. Berlin argues—and the proposal was recently enacated into law as the “Making Work Pay Tax Credit”— that rising earnings from an expanded EITC might therefore deliver benefits in addition to direct income, including higher marriage rates, decreased relative attractiveness of illicit sources of income, and so on.52 In addition, Daniel P. Gitterman, Lucy S. Gorham, and Jessica L. Dorrance (2007) suggest that an expanded EITC for individuals without qualifying children would also recognize that many noncustodial parents have responsibility for children, especially in light of high out-of-wedlock childbearing and divorce rates.

62

/

Alternative Labor-Market Policies

There is evidence that declining wages for less-skilled men have reduced employment (for example, Juhn 1992). Consistent with this evidence, Berlin (2007) cites evidence from experimental evaluations of programs offering earnings supplements to low-wage workers conditional on working thirty hours or more per week, which points to positive effects on earnings (net of the income supplement). However, most of the evidence pertains to single mothers. One program (the New Hope community group in Milwaukee) offered supplements to single men, and it generally produced positive effects on employment, earnings, and family income of men, even up to five years after the program, although many of the estimated effects are not significant (Duncan et al. 2007). However, regarding the broader link between higher earnings and the encouragement of marriage and discouragement of crime, the evidence is certainly more sparse. Scholz (2007), who also advocates making the EITC for unrelated individuals more generous, discusses some of the research linking lower wages to higher crime and lower marriage rates, while acknowledging that this evidence is fairly limited, especially in the case of marriage. Finally, there appears to be no evidence on how an expanded EITC for individuals might translate into more resources for children of noncustodial parents. Clearly, though, if such a policy would lead to major reductions in crime, for example, then the benefits could outweigh the costs, as Scholz (2007) argues. However, a number of issues arise in considering this policy. To begin with, it is useful to think about how the EITC affects wages and the margins on which it operates in order to try to clarify who might gain and lose from an expanded EITC. The current EITC boosts employment among those who would not work in its absence, which should increase competition with those already in the labor market and reduce the market wage for low-skilled workers.53 Indeed, Jesse Rothstein (2007) reports such evidence. Aside from implying that employers get some of the gains from an expanded EITC, this evidence also points to some important considerations with regard to expanding the EITC for unrelated individuals. In particular, if expanding the EITC for unrelated individuals brings more such individuals into the labor market—which is part of the argument (Berlin 2007; Gitterman et al. 2007)—then this would lower low-skilled wages, potentially shifting some of the benefits of the EITC program as a whole away from families, as the eligible participants with children face increased labor supply from unrelated individuals, while at the same time encouraging employment and increasing earnings and income among less-skilled men who are not in families. There may be an argument for shifting the benefits of the EITC in this way, but if so this should be made explicit, rather than suggesting that we can simply do more to increase incomes at the bottom of the distribution by extending the EITC to unrelated individuals, with no tradeoffs. On the other hand, if it can be established that an expanded EITC for unrelated individuals has substantial impacts on the employment margin, the potential benefits from returns to labor-market experience that eventually lead to increased earnings and reduced reliance on the EITC might prove a substantial boon to economic self-sufficiency.54 Also, if the unrelated individuals who are the target of proposals to expand the EITC are already working, an expanded

/

63

Making the Work-Based Safety Net Work Better

EITC would likely reduce their labor supply but increase their total income (earnings plus EITC), and this reduction in labor supply would increase low-skilled wages, likely benefiting other EITC recipients as well. It is also useful to think about the main goals of expanding the EITC for unrelated individuals, and asking whether doing so is likely to be the most efficient policy option. In particular, to the extent that social goals relating to marriage and child support figure prominently in proposals for a more generous EITC for unrelated individuals, it may make more sense to try to reduce directly some of the incentives of the EITC with respect to marriage and to pursue other measures to increase support of unmarried parents for children, as it seems at least possible that “buying” changes in these behaviors through an expanded EITC for unrelated individuals might prove very inefficient. The same argument could be made about crime, although I admit that arguments that we need to do more to increase the costs of crime, rather than trying to increase the benefits of participating in the licit economy, seem far-fetched in the current U.S. context of massive incarceration. Aside from the idea of extending the EITC to unrelated individuals, there have also been proposals to structure the EITC differently, to focus on low-wage workers (MaCurdy 2004). As noted earlier, part the motivation for this structure is to try to reduce the labor supply disincentives associated with the EITC for higherincome families that are still eligible. But this idea is also intended to better target families with low-wage workers that cannot, even with full-time work, achieve an acceptable standard of living, and to increase work incentives. Thomas MaCurdy discusses two options: a “wage-based” EITC that pays a share of the maximum EITC benefit based on the share of full-time work in a family; and a “wage-subsidy” EITC that pays the difference between a target wage and the worker’s market wage (when it is lower), multiplied by hours worked (averaged for the family). In the context of state add-ons to the federal EITC (in data for California), he finds that the wage-based policy provides similar incentives and benefits for low-wage workers, but greater work incentives for higher-wage workers, while lowering expenses by reducing benefits for families with high-wage and part-time workers, and that the wage-subsidy policy targets families with low-wage workers, and provides stronger work incentives. As MaCurdy notes, however, these alternative EITC programs pose greater administrative challenges, because of the necessity of measuring family labor supply. Nonetheless, these ideas for modifying the EITC merit further consideration because of their potential for boosting earnings among low-income families. Finally, a new but growing literature on school-to-work provides some support for the potential benefits of school-to-work institutions and programs. Although there is an absence of evidence on longer-run effects, it appears that institutions and programs to improve the school-to-work transition deliver benefits in terms of labor-market attachment, skill formation, and higher wages and earnings. However, there are a number of missing pieces of evidence, in particular with regard to the longer-term effects of specific programs. In the past decade, school-towork efforts were largely dismantled, in favor of test-based educational reforms. The evidence suggests that explicit school-to-work programs—such as those en-

64

/

Alternative Labor-Market Policies

couraged by the STWOA, as well as career academies—deliver benefits. This suggests that policymakers concerned with improving education to enhance labormarket success would likely do better by pursuing a more balanced mix of strengthening school-to-work institutions and programs along with other educational reforms than has been the case with the either-or approach of federal policy in the last decade and a half.55 At the same time, there are two qualifications to this suggestion simply to pursue both types of policies. First, under any circumstances there will be trade-offs between school-to-work efforts and a greater focus on academic preparation, as there is a fixed amount of time that students can allocate during the school day. Indeed, concerns that NCLB has cut into the time available for course electives generally, and work-related courses specifically, have already been voiced (see, for example, Stone and Aliaga 2007, 82).56 And second, it is essential to remain cognizant of the limitation of school-to-work programs as well as education reform efforts, which serve current students who will work in the future, but not older individuals already in the labor market.57 Because I have strived to discuss the evidence on policies to increase economic self-sufficiency via higher earnings, I have of necessity focused on policies that are already in place (or were in place recently). All these policies focus on mandating higher wages, encouraging work, or increasing skills. I have suggested that we have (or had) in place successful policies to encourage work (the EITC) and to increase skills (school-to-work). On the other hand, I have argued that mandating higher wage floors is ineffective or even counterproductive. An obvious question is whether any new policies that have been tried on a much less extensive basis might hold promise. One I discussed in some detail is the expansion of work subsidies (perhaps in the form of the EITC) to unrelated individuals, although I have suggested that this may have undesirable distributional effects. Focusing instead on skill formation, Harry J. Holzer (2007) has advocated a broad-based system of federal grants to encourage skill formation, which he labels Worker Advancement Grants for Employment in States (WAGES). The call for this broad-based program rests in part on fairly compelling evidence of beneficial effects (such as for school-to-work programs or community college enrollment), and in part on less rigorous evidence pointing to isolated examples of other types of programs that appear to be particularly effective, even if they are less effective in other locations or settings. Recognizing that the research evidence on the effects of these other programs is weak, Holzer has proposed simultaneous adoption of a competitive process for states to receive federal funds, based on program evaluation. There are legitimate questions as to whether building evaluation requirements into the program will be effective in creating a process whereby the most effective programs emerge and crowd out the least effective.58 Moreover, I think it is likely that Holzer’s back-of-the envelope calculations about the social returns to investments in these programs are overoptimistic.59 Nonetheless, it is probably true that in the long run, policies that increase skills are the Holy Grail that can give individuals the capacity to achieve economic selfsufficiency through their own earnings, and that we should be open to some of

/

65

Making the Work-Based Safety Net Work Better

the new ideas (labor-market intermediaries, sectoral training, career ladders, and so on) that Holzer discusses. Especially in light of what is probably a cultural bias in the United States toward meeting the goals of economic self-sufficiency through earnings rather than public support, it is hard to argue against the merit, in principle, of efforts devoted to trying to raise skills of the less-skilled and less-advantaged population. And if a structure can be successfully built that encourages innovative programs while at the same time being hard-nosed about program evaluation—discarding unsuccessful programs, promoting the successful ones, and killing the whole project if it does not deliver programs that are cost-effective and scalable—then we may ultimately learn something useful and make some progress regarding policies to raise skills and increase the economic self-sufficiency of individuals and families through labor-market earnings.

This chapter was prepared for the conference “Pathways to Self Sufficiency: Getting Ahead in an Era beyond Welfare Reform,” held in Madison, Wisconsin, September 6 to 7, 2007. I am grateful to Greg Duncan, Carolyn Heinrich, Harry Holzer, Karl Scholz, other conference participants, and two anonymous reviewers for helpful comments. The views expressed are not those of the Public Policy Institute of California.

NOTES 1. Of course shifts in aggregate demand will have the same effect, but my focus is on labor-market policies. 2. Aside from what the government can do, there is a large and growing literature on the effects on firms and workers of what are often called “high-road” work practices on firms and workers that lead to more productive and higher-wage jobs (see, for example, Cappelli and Neumark 2001; Handel and Levine 2004). However, although policies to encourage firms to adopt such practices are often advocated (for example, Osterman et al. 2002), it is unclear exactly what policies would achieve this. One argument that is sometimes made is that a higher minimum wage would encourage employers to adopt such practices, by raising the productivity level at which it is profitable to hire a worker (Bernstein 2000; Fitzgerald 2006), in which case high-road practices might lead to skill upgrading. Of course a higher minimum wage increases the productivity of labor in the neoclassical model with no necessary implications for what practices firms adopt, as firms move up the labor demand curve; and there is no direct evidence that minimum wages change firms’ practices. Another tactic that has emerged is corporate responsibility campaigns by community groups or labor unions to encourage the creation of high-wage jobs; see, for example, http://www.working foramerica.org/documents/Journal4/regional.htm (accessed May 2, 2007). 3. On welfare reform, see, for example, Rebecca M. Blank (2002); regarding training, see, 66

/

Alternative Labor-Market Policies

4. 5. 6.

7. 8. 9. 10. 11.

12.

13. 14. 15. 16 17. 18.

19.

for example, James J. Heckman, Robert J. Lalonde, and Jeffrey A. Smith (1999). For a very recent discussion of adult training and other workforce development policies, see Holzer and Nightingale (2007). In addition, Marianne P. Bitler, Jonah B. Gelbach, and Hilary W. Hoynes (2006) point out the importance of heterogeneity in these effects. See http://www.uses.doleta.gov/wotcdata.cfm (accessed May 2, 2007). For evidence on efforts to increase schooling levels at the postsecondary level, see, for example, Christopher Cornwell, David B. Mustard, and Deepa J. Sridhar (2006). One policy effort to increase primary or secondary schooling of disadvantaged youths is the “learnfare” program, implemented in a handful of states. Under learnfare, welfare benefits can be cut if children of recipient families are not in school. See http://www.dol.gov/esa/minwage/america.htm (accessed May 2, 2007). See Card and Krueger (1994), and also Neumark and Wascher (2000). See John M. Abowd et al. (2000), Janet Currie and Bruce Fallick (1996), and Neumark, Schweitzer, and Wascher (2004). The review also discusses in some detail our efforts to explain some of the conflicting findings in the literature. Since this review was published, additional research on both sides of the debate has continued to appear. A couple of provocative papers that continue to challenge the conclusion that minimum wages reduce employment among the least-skilled have been written by Sylvia Allegretto, Arindrajit Dube, and Michael Reich (2008) and Arindrajit Dube, T. William Lester, and Michael Reich (2008). For example, in response to a presentation of the findings from the minimum wage review just discussed, Jared Bernstein of the Economic Policy Institute, which advocates a much higher minimum wage, responded, “The minimum wage increase will invariably hurt some of its intended beneficiaries . . . [but] the benefits will often outweigh the costs, even for narrowly-affected workers.” See http://www.aei.org/events/ eventID.1430,filter.all/event_detail.asp# (accessed June 11, 2007). See Neumark and Wascher (2002) for a more detailed discussion and illustrative calculations. The same arguments would apply to hours effects. The sample conditions on the initial wage, and hence the models, are estimated for those initially working. See Neumark, Schweitzer, and Wascher (2004, figure 1). Near-poor families are defined as those at 1.0 to 1.5 times the poverty line. For the full set of figures, see figure 1 of Neumark, Schweitzer, and Wascher (2005). A more detailed discussion is provided in Neumark and Wascher (2008). I omit from table 2.3 a study by John T. Addison and McKinley L. Blackburn (1999) that focuses only on narrow subsets of families and therefore does not speak to the overall effects of minimum wages on poverty or the income distribution. There are some related questions about the effects of minimum wages on the family income distribution for different types of families. For example, minimum wages may have different effects on rural and urban families. Wu, Perloff, and Golan (2006b) report evidence suggesting that minimum wages increase pre-tax and post-tax family income inequality in urban areas, and more so when more weight is put on the lower end of the distribution. These results for urban areas are consistent with their aggregated results, presented in table 2.3. However, they find very small and insignificant /

67

Making the Work-Based Safety Net Work Better

20.

21.

22. 23.

24.

68

distributional effects of minimum wages in rural areas. Craig Gunderson and James P. Ziliak (2004) contrast some of their results for female-headed versus married-couple families. Across their different specifications there is no clear pattern of differences across family structure in how the minimum wage affects the poverty rate or the squared poverty gap, and none of the estimates by family structure are significant. The most recent development with regard to living wages has been the advent of citylevel minimum wages—that is, broad minimum-wage floors just like state minimum wages, but enacted at the city level. Santa Fe and San Francisco enacted a minimum of $8.50 in 2003 and 2004, respectively, with both set to rise through indexation and (in Santa Fe) additional planned increases in the legislation. Madison and other smaller cities in Wisconsin also recently passed minimum-wage laws, but they were subsequently repealed by state laws. A city minimum wage in New Orleans was approved by voters in 2002, but subsequently was blocked by a state law. (Washington, D.C., has its own minimum wage, but the District is often treated as a state in state-level analyses of minimum-wage effects like those described earlier.) See the summary of coverage estimates in Adams and Neumark (2005a) and in Richard Freeman (2005). For more systematic estimates of coverage by the living wage laws in Los Angeles and San Francisco, respectively, see David Fairris et al. (2005) and Susan Alunan et al. (1999). A very recent review of the living wage literature, which also discusses some other studies, is provided in Holzer (2008). It might be viewed as curious that the estimate for contractor-only living-wage laws, although insignificant, is larger, in light of the smaller wage and employment effects for contractor-only laws reported in columns 1 and 2. But the offsetting positive wage effects and negative employment effects of financial-assistance living-wage laws imply that these laws need not have a stronger effect on poverty. Nonetheless, the estimate for contractor-only laws is puzzling in light of the absence of wage or employment effects of these narrower laws. The size of some of these estimated effects may seem surprisingly large, given relatively low coverage by living-wage laws. With respect to poverty reductions, Adams and Neumark (2004) explain that their estimates are of the same order of magnitude suggested by Mark Brenner’s (2005) calculations based on detailed data for Boston. With respect to wage and employment effects, the estimates are large, given coverage estimates. One possibility is that there are non-neoclassical influences, so that livingwage laws affect community norms for wages; one channel for this may be that firms desirous of future contracts or development subsidies believe it is advantageous to pay higher wages (Bartik 2004). At the same time, some critics have grossly overstated the implications of these estimates for employment declines. Fairris and Michael Reich (2005, 10) incorrectly calculate that the Adams and Neumark estimates imply huge employment losses. In fact, the 35 percent living-wage increase that they consider, for a financial-assistance living-wage law, is estimated to lead to a 6 percent employment decline among those in the bottom tenth of the skill distribution (.35 multiplied by the employment effect of –.076 reported in column 2 of table 2.5, divided by the .44 employment rate in the bottom tenth of the skill distribution). This contrasts with the 91 percent employment decline that Fairris and Reich claim is implied by the estimates.

/

Alternative Labor-Market Policies 25. Teenagers may also be myopic, putting too much emphasis on the potential higher earnings in the minimum-wage sector following a minimum-wage increase, with insufficient attention to the cost in terms of foregone higher earnings from more education. 26. As confirmatory evidence, this negative effect was present for observations in states with compulsory schooling ages less than eighteen (where teenagers have more choice about leaving school), but not in states with a compulsory schooling age of eighteen (for which the estimates were smaller and insignificant, although still negative). 27. These parameters determine the level of income at which the credit falls to zero, which was $36,348 in 2006, for a family with two children. 28. The nineteen states with EITC supplements in 2007 were Delaware, Illinois, Indiana, Iowa, Kansas, Maine, Maryland, Massachusetts, Minnesota, Nebraska, New Jersey, New Mexico, New York, Oklahoma, Oregon, Rhode Island, Vermont, Virginia, and Wisconsin, and the supplemental EITC in those states ranged from 4 to 43 percent. See Neumark and Wascher (2009). In addition, EITC supplements became effective in 2008 in Louisiana, Michigan, and North Carolina. 29. The usual substitution effect occurs as the higher wage raises the price of leisure and induces more labor supply, in the case of those originally nonemployed, leading some to enter the labor market. The income effect, however—which typically induces a labor supply reduction as a result of the higher wage leading to more consumption of all normal goods, including leisure—does not have an impact on those originally working zero hours. 30. A rare exception that fails to find employment (or hours) effects is Cancian and Levinson (2005). 31. For this reason, Thomas MaCurdy (2004) considers EITC alternatives that target lowwage workers. These alternatives are discussed more in the concluding section. 32. This evidence emerges in specifications that rely on the state-level variation in EITC parameters; the paper argues that federal variation is confounded with other policy changes. Jeffrey Grogger (2003) reports confirming evidence, finding that higher EITC payments increase employment, hours, and earnings of female-headed households. 33. This effect is perhaps a little implausibly large. As reported in the paper, the estimated effect tapers off considerably with increased generosity. 34. Wu, Perloff, and Golan (2006a) conclude that higher marginal tax rates have larger beneficial redistributional effects than the EITC. They find adverse effects of AFDC and Food Stamps, which they attribute to work disincentives. They also find that a higher phase-out rate increases inequality for nearly every inequality measure, presumably because a higher phase-out rate reduces labor supply among moderateincome families. Craig Gunderson and James P. Ziliak (2004) find mixed evidence on the distributional effects of the EITC. Their evidence is more consistent with the EITC increasing the poverty rate, using the before- or after-tax measure (the latter incorporates the credit), although only some of the estimates are statistically significant. For the squared poverty gap, with the before-tax measure there are no significant effects of the EITC, whereas for the after-tax measure there is some evidence consistent with a reduction in the gap, although curiously this emerges for married-couple families,

/

69

Making the Work-Based Safety Net Work Better

35.

36.

37. 38.

39.

40.

41.

42.

43.

70

which is unexpected. Their specification may be problematic because it does not appear to allow the effects of the EITC and other policies to vary with the number of children in the family, which in Neumark and Wascher (2001b) appeared to be quite important. They use the difference between the logs of the state and federal maximum EITC benefit, although it is not clear for what type of family, or whether this is averaged over families in each state-year cell. And Dickert-Conlin and Holtz-Eakin (2000) present evidence that offering wage subsidies for low-skilled workers is also not particularly effective at targeting poor families. The NJTC targeted new employment, not specific workers, but it created stronger incentives to hire low-wage workers by applying only to the first $4,200 of wages per employee (in 1977 and 1978). There are also similar efforts tied to welfare reform at the federal level, with the Welfare to Work Tax Credit. However, Sarah Hamersma (2005a) summarizes a 2001 GAO report suggesting little evidence of such churning, as very few workers receiving subsidies stayed long enough to get the maximum subsidy, and hiring and training costs appeared to render such a strategy cost ineffective. Many researchers have noted that most EITC recipients choose to take their payment as a lump sum at the end of the year rather than in each paycheck. It is conceivable that part of the explanation for this is the avoidance of stigma effects. James J. Heckman, Robert J. Lalonde, and Jeffrey A. Smith (1999) provide an extensive review of the literature documenting this general result, based on data for both the United States and Europe. (See also the brief summary in Holzer [2007], who concurs with this summary of the findings from the literature on training programs.) Heckman, Lalonde, and Smith summarize cost-benefit calculations indicating that training programs for adults sometimes appear to deliver positive social returns. Recent evidence from evaluations of the Job Corps program appear to indicate gains for twentyto twenty-four-year-olds, but not those younger (see Alan Krueger, “New (and Sometimes Conflicting) Data on the Value to Society of the Job Corps,” New York Times, January 5, 2006. Available at: http://www.nytimes.com/2006/01/05/business/05scene .html?_r=1&oref=slogin [accessed January 10, 2008]). The federal funding was intended to serve as seed money to establish school-to-work transition systems that included formal partnerships among secondary and postsecondary institutions and employers. Research has established that in many states the legislation did spur the development of such systems (see, for example, Hershey et al. 1999; Neumark and Allen 2003; Neumark 2006). In contrast, vocational education (later called career and technical education) was characterized by its isolation from more comprehensive high school curricula (Hayward and Benson 1993), whereas Tech Prep pursued limited integration of vocational and academic education by introducing some vocational education into comprehensive high school curricula while sequencing vocational education during high school with two years of further related study at postsecondary institutions (Ryan 2001). As explained in Neumark and Rothstein (2006), these expectations variables can also be interpreted as ideal proxies, the inclusion of which may eliminate bias from en-

/

Alternative Labor-Market Policies

44.

45.

46.

47.

48. 49. 50.

51.

52.

dogenous selection into school-to-work participation. The idea is that any remaining variation in unobservables net of these expectations is only forecast error, which should be uncorrelated with the exogenous variables in the information set. The link between school enterprise and college enrollment is not obvious. However, research by David Stern (1984) and Stern et al. (1994) reports that enterprise-based jobs are more closely related to students’ education than are out-of-school jobs, and that enterprise-based jobs provide more opportunity to apply what students are learning in school; so school enterprises may be a particular type of school-to-work program that enhances the educational experience and therefore encourages higher education. We found that very few variables were predictive of participation in school-to-work programs; the one important exception was that blacks were more likely to participate in coop, school enterprise, and Tech Prep programs. These results suggest that there may be little systematic selection into school-to-work programs, which would explain why the results are insensitive to adding controls. Of course this share is not necessarily one-half of the relevant population. But the “forgotten half” phrase was introduced by the William T. Grant Foundation (1988), and this characterization has stuck in subsequent work (see, for example, Donahoe and Tienda 1999). Kemple (2005) reports that there are approximately 2,500 career academies across the country. He does not report any direct enrollment numbers, but suggests that academies typically enroll 150 to 200 students. These numbers suggest that there are around 440,000 students in career academies, out of approximately 16 million high school students. The latter number comes from Digest of Education Statistics, available at: http://nces.ed.gov/programs/digest/d04/tables/dt04_002.asp (accessed May 14, 2007); Kemple does not give a date (or source) for his estimate. The National Center for Education Statistics (NCES) reports 4,800 secondary schools with career academies, nearly twice Kemple’s estimates; see http://nces.ed.gov/surveys/ctes/tables/ h14.asp (accessed May 14, 2007). However, it is not clear what the differences are in the definitions of career academies. The study covers nine schools across the country, all located in or near urban areas. These results appear to be for men and women combined. Note that this evidence for men contrasts with the findings in Orr et al. (2007) pointing to positive effects of the National Academy Foundation (NAF) career academies on four-year college attendance. It is unclear how much this difference is attributable to the random assignment or to differences in the types of academies studied. In particular, Orr et al. suggest that the NAF academies cover particular fields (especially finance, which was the type that boosted college attendance), and may differ in terms of other program characteristics as well, given that the academies they study were relatively homogeneous and met criteria for implementation of the NAF model. John Karl Scholz proposes targeting lower-income families with this subsidy by restricting eligibility to individuals living in federally designated Renewal Communities, Empowerment Zones, or Enterprise Communities. However, his analysis suggests that a fairly high share of the gains would go to families above the poverty line. See http://johnedwards.com/about/issues/poverty (accessed June 26, 2007).

/

71

Making the Work-Based Safety Net Work Better 53. It is also worth pointing out, in relating the literature on minimum-wage employment effects to the EITC, that the less elastic one thinks the demand for low-skilled labor is, the more one would expect wages for low-skilled workers to fall in response to an expanded EITC. 54. This is speculative, but is in some sense the flip side of the results in Neumark and Nizalova (2007) suggesting that minimum wages may have long-term adverse effects on wages and employment, stemming in part from lost work experience. 55. Although difficult to establish rigorously, it is also my impression that the $1.5 billion in federal funds for the STWOA seem to have had rather dramatic policy impacts in terms of spurring the creation of school-to-work programs in the states, presumably because the seed money provided by the STWOA interacted with other incentives for states to create these programs. 56. David Figlio (2007) discusses a number of studies on the issue of trade-offs between resources devoted to accountability programs and other activities in schools. 57. This qualification is even more relevant to discussions of the potential higher efficacy of early (that is, childhood) human capital investment (for example, Carneiro and Heckman 2002). 58. Neumark (2006) presents an analysis of program evaluation of school-to-work programs in California that raises warning flags about the implementation of evaluation mandates. On the other hand, Blank (2002) argues that the process of experimentation and evaluation was very successful in the arena of welfare reform. 59. In particular, his calculations are based on estimates of gains from programs that are apparently the highest among the estimates he surveys, and in each case come from an isolated program when estimates of the earnings gains from similar programs implemented elsewhere were smaller and often insignificant (see Holzer 2007, 20). Whether these high estimates reflect best practices or simply statistical outliers, it is probably unreasonable to presume that gains of these magnitudes will be replicated across the board, even with the best institutional structure in place.

REFERENCES Abowd, John M., Francis Kramarz, David N. Margolis, and Thomas Phillipon. 2000. “The Tail of Two Countries: Minimum Wages and Employment in France and the United States.” Discussion paper no. 203. Bonn, Germany: Institute for the Study of Labor (IZA). Acemoglu, Daron, and Jörn-Steffen Pischke. 2003. “Minimum Wages and On-the-Job Training,” Research in Labor Economics 22: 159–202. Adams, Scott, and David Neumark. 2004. “The Economic Effects of Living Wage Laws: A Provisional Review.” Urban Affairs Review 40(2): 210-45. ———. 2005a. “When Do Living Wages Bite?” Industrial Relations 44(1): 164–92. ———. 2005b. “Living Wage Effects: New and Improved Evidence.” Economic Development Quarterly 19(1): 80–102. Addison, John T., and McKinley L. Blackburn. 1999. “Minimum Wages and Poverty.” Industrial and Labor Relations Review 52(3): 393–409. Allegretto, Sylvia, Arindrajit Dube, and Michael Reich. 2008. “Do Minimum Wages Really Re-

72

/

Alternative Labor-Market Policies duce Teen Employment? Accounting for Heterogeneity and Selectivity in State Panel Data.” Working paper no. 166-08. Berkeley, Calif.: Institute for Research on Labor and Employment. Alunan, Susan, Lisel Blash, Brian Murphy, Michael J. Potepan, Hadley Roff, and Odilla Sidime-Brazier. 1999. The Living Wage in San Francisco: Analysis of Economic Impact, Administrative, and Policy Issues. San Francisco: San Francisco Urban Institute. Atkinson, Anthony B. 1970. “On the Measurement of Inequality.” Journal of Economic Theory 2(3): 244–63. Bartik, Timothy J. 2004. “Thinking About Local Living Wage Requirements.” Urban Affairs Review 40(2): 269–99. Berlin, Gordon L. 2007. “Rewarding the Work of Individuals: A Counterintuitive Approach to Reducing Poverty and Strengthening Families.” The Future of Children 17(2): 17–42. Bernstein, Jared. 2000. “Two Cheers for the EITC.” American Prospect Online, June 19. Available at: http://www.prospect.org/web/page.ww?section=root&name=ViewPrint&arti cleId=4331 (accessed May 2, 2007). Betts, Julian R. 1995. “Does School Quality Matter? Evidence from the National Longitudinal Survey of Youth.” Review of Economics and Statistics 77(2): 231–50. Bitler, Marianne P., Jonah B. Gelbach, and Hilary W. Hoynes. 2006. “What Mean Impacts Miss: Distributional Effects of Welfare Reform Experiments.” American Economic Review 96(4): 988–1012. Blank, Rebecca M. 2002. “Evaluating Welfare Reform in the United States.” Journal of Economic Literature 40(4): 1105–66. Blank, Rebecca M., and Lucie Schmidt. 2001. “Work, Wages, and Welfare.” In The New World of Welfare, edited by Rebecca M. Blank and Ron Haskins. Washington, D.C.: Brookings Institution. Brenner, Mark. 2005. “The Economic Impact of the Boston Living Wage Ordinance.” Industrial Relations 44(1): 59–83. Brown, Charles, Curtis Gilroy, and Andrew Kohen. 1982. “The Effect of the Minimum Wage on Employment and Unemployment.” Journal of Economic Literature 20(2): 487–528. Burkhauser, Richard V., Kenneth A. Couch, and David C. Wittenburg. 1996. “‘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 49(3): 547–52. Burkhauser, Richard V., and Joseph J. Sabia. 2007. “The Effectiveness of Minimum-Wage Increases in Reducing Poverty: Past, Present and Future.” Contemporary Economic Policy 25(2): 262–81. Cancian, Maria, and Arik Levinson. 2005. “Labor Supply Effects of the Earned Income Tax Credit: Evidence from Wisconsin’s Supplemental Benefit for Families with Three Children.” NBER Working Paper No. 11454. Cambridge, Mass.: National Bureau of Economic Research. Cappelli, Peter, and David Neumark. 2001. “Do “High-Performance” Work Practices Improve Establishment-Level Outcomes?” Industrial and Labor Relations Review 54(4): 737–75. Card, David. 1992. “Do Minimum Wages Reduce Employment? A Case Study of California, 1987–1989.” Industrial and Labor Relations Review 46(1): 38–54.

/

73

Making the Work-Based Safety Net Work Better Card, David, and Alan B. Krueger. 1992. “School Quality and Black-White Relative Earnings: A Direct Assessment.” Quarterly Journal of Economics 107(1): 151–200. ———. 1994. “Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania.” American Economic Review 84(5): 772–93. ———. 1995. Myth and Measurement: The New Economics of the Minimum Wage. Princeton, N.J.: Princeton University Press. Carneiro, Pedro, and James J. Heckman. 2002. “Human Capital Policy.” In Inequality in America, edited by James J. Heckman and Alan B. Krueger. Cambridge, Mass.: MIT Press. Chaplin, Duncan D., Mark D. Turner, and Andreas D. Pape. 2003. “Minimum Wages and Skill Acquisition: Another Look at Schooling Effects.” Economics of Education Review 22(1): 11–22. Cohen, Marie, and Douglas J. Besharov. 2002. “The Role of Career and Technical Education: Implications for the Federal Government.” Unpublished paper. Washington, D.C.: U.S. Department of Education, Office of Vocational and Adult Education. Commission on the Skills of the American Workforce. 1990. America’s Choice: High Skills or Low Wages. Rochester, N.Y.: National Center on Education and the Economy. Cornwell, Christopher, David B. Mustard, and Deepa J. Sridhar. 2006. “The Enrollment Effects of Merit-Based Financial Aid: Evidence from Georgia’s HOPE Program.” Journal of Labor Economics 24(4): 761–86. Cunningham, James. 1981. “The Importance 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, D.C.: American Enterprise Institute. Currie, Janet, and Bruce Fallick. 1996. “The Minimum Wage and the Employment of Youth: Evidence from the NLSY.” Journal of Human Resources 31(2): 404–28. Dickert-Conlin, Stacy, and Douglas Holtz-Eakin. 2000. “Employee-Based Versus EmployerBased Subsidies to Low-Wage Workers: A Public Finance Perspective.” In Finding Jobs: Work and Welfare Reform, edited by David E. Card and Rebecca M. Blank. New York: Russell Sage Foundation. Donahoe, Debra, and Marta Tienda. 1999. “Human Asset Development and the Transition from School to Work: Policy Lessons for the 21st Century.” Unpublished paper. Princeton, N.J.: Princeton University, Office of Population Research. Dube, Arindrajit, T. William Lester, and Michael Reich. 2008. “Minimum Wage Effects Across State Borders: Estimates Using Contiguous Counties.” Working paper no. 157-07. Berkeley, Calif.: Institute for Research on Labor and Employment. Duncan, Greg J., Heather D. Hill, and Aleksy Tetenov. 2007. “The Persistence of New Hope’s Labor Market Impacts: How Long? How Real?” Unpublished paper. Evanston, Ill.: Northwestern University. Ehrenberg, Ronald G., and Alan J. Marcus. 1980. “Minimum Wage Legislation and the Educational Outcomes of Youths.” Research in Labor Economics 3: 61–93. Fairris, David, and Roberto Pedace. 2004. “The Impact of Minimum Wages on Job Training: An Empirical Exploration with Establishment Data.” Southern Economic Journal 70(3): 566–83. Fairris, David, and Michael Reich. 2005. “The Impacts of Living Wage Policies: Introduction to the Special Issue.” Industrial Relations 44(1): 1–13.

74

/

Alternative Labor-Market Policies Fairris, David, David Runsten, Carolina Briones, and Jessica Goodheart. 2005. Examining the Evidence: The Impact of the Los Angeles Living Wage Ordinance on Workers and Businesses. Available at: http://www.losangeleslivingwagestudy.org (accessed November 6, 2006). Feldstein, Martin. 1973. “The Economics of the New Unemployment.” Public Interest 33(Fall): 3–42. Figlio, David N. 2007. “School Reforms and Low-Income Families.” Unpublished paper. Gainesville: University of Florida. Fitzgerald, Joan. 2006. “Getting Serious About Good Jobs.” The American Prospect, October 22. Available at: http://www.prospect.org/cs/articles?article=getting_serious_about _good_jobs (accessed June 25, 2007). Freeman, Richard B. 1996. “The Minimum Wage as a Redistributive Tool.” The Economic Journal 106(436): 639–49. ———. 2005. “Fighting for Other Folks’ Wages: The Logic and Illogic of Living Wage Campaigns.” Industrial Relations 44(1): 14–31. Gitterman, Daniel P., Lucy S. Gorham, and Jessica L. Dorrance. 2007. “Expanding the EITC for Single Workers and Couples Without Children (aka Tax Relief for Low-Wage Workers).” Unpublished paper. Chapel Hill: University of North Carolina. Gramlich, Edward M. 1976. “Impact of Minimum Wages on Other Wages, Employment, and Family Incomes.” Brookings Papers on Economic Activity 7(2): 409–51. Grogger, Jeffrey. 2003. “The Effects of Time Limits, the EITC, and Other Policy Changes on Welfare Use, Work, and Income Among Female-Headed Households.” Review of Economics and Statistics 85(2): 394–408. Grubb, W. Norton. 2001. “Second Chances in Changing Times: The Roles of Community Colleges in Advancing Low Wage Workers.” In Low Wage Workers in the New Economy, edited by Richard Kazis and Marc Miller. Washington, D.C.: Urban Institute Press. Gunderson, Craig, and James P. Ziliak. 2004. “Poverty and Macroeconomic Performance Across Space, Race, and Family Structure.” Demography 41(1): 61–86. Hamersma, Sarah. 2005a. “The Work Opportunity and Welfare-to-Work Tax Credits.” Urban-Brookings Tax Policy Center Brief no. 15. Washington, D.C.: Urban Institute, October. ———. 2005b. “The Effects of an Employer Subsidy on Employment Outcomes: A Study of the Work Opportunity and Welfare to Work Tax Credits.” Institute for Research on Poverty Discussion Paper 1303-05. Madison: University of Wisconsin. Hamilton, Stephen F. 1990. Apprenticeship for Adulthood. New York: Free Press. Handel, Michael J., and David I. Levine. 2004. “Editors’ Introduction to Special Issue: The Effects of Work Practices on Workers.” Industrial Relations 43(1): 1–43. Hanushek, Eric A. 2006. “Alternative School Policies and the Benefits of General Cognitive Skills.” Economics of Education Review 25(4): 430–46. Hashimoto, Masanori. 1982. “Minimum Wage Effects on Training on the Job.” American Economic Review 72(5): 1070–87. Hayward, G. C., and C. S. Benson. 1993. “The Changing Role of Vocational-Technical Education in the United States.” CenterWork (National Center for Research on Vocational Education) 4(2): 1–3. Heckman, James J., Robert J. Lalonde, and Jeffrey A. Smith. 1999. “The Economics and Econometrics of Active Labor Market Programs.” In Handbook of Labor Economics, edited by Orley Ashenfelter and David Card. Volume 3. Amsterdam: Elsevier Science B.V.

/

75

Making the Work-Based Safety Net Work Better Hershey, Alan M., Marsha K. Silverberg, and Joshua Haimson. 1999. Expanding Options for Students: Report to Congress on the National Evaluation of School-to-Work Implementation. Princeton, N.J.: Mathematica Policy Research. Hoffman, Saul D., and Laurence S. Seidman. 2003. Helping Working Families: The Earned Income Tax Credit. Kalamazoo, Mich.: W.E. Upjohn Institute. Holzer, Harry J. 2007. “Better Workers for Better Jobs: Improving Worker Advancement in the Low-Wage Labor Market.” Hamilton Project Discussion Paper. Washington, D.C.: Brookings Institution. ———. 2008. “Living Wage Laws: How Much Do (Can) They Matter?” Unpublished paper. Washington, D.C.: Georgetown University. Holzer, Harry J., and Demetra Smith Nightingale, eds. 2007. Reshaping the American Workforce in a Changing Economy. Washington, D.C.: Urban Institute Press. Horrigan, Michael W., and Ronald B. Mincy. 1993. “The Minimum Wage and Earnings and Income Inequality.” In Uneven Tides: Rising Inequality in America, edited by Sheldon Danziger and Peter Gottschalk. New York: Russell Sage Foundation. Hotz, V. Joseph, and John Karl Scholz. 2003. “The Earned Income Tax Credit.” In MeansTested Transfer Programs in the United States, edited by Robert A. Moffitt. New York: Russell Sage Foundation. Hughes, Katherine L., Thomas R. Bailey, and Melinda J. Mechur. 2001. School-to-Work: Making a Difference in Education. New York: Columbia University, Teachers College, Institute on Education and the Economy. Juhn, Chinhui. 1992. “Decline of Male Labor Market Participation: The Role of Declining Market Opportunities.” Quarterly Journal of Economics 107(1): 79–121. Katz, Lawrence F. 1998. “Wage Subsidies for the Disadvantaged.” In Generating Jobs: How to Increase Demand for Less-Skilled Workers, edited by Richard B. Freeman and Peter Gottschalk. New York: Russell Sage Foundation. Kemple, James J. 2001. Career Academies: Impacts on Students’ Initial Transitions to PostSecondary Education and Employment. New York: Manpower Demonstration Research Corporation. ———. 2004. “Making an Impact on the Transition from School to Work: Evidence from Career Academies.” Unpublished paper. New York: Manpower Demonstration Research Corporation. Kemple, James J., and Jason C. Snipes. 2000. Career Academies: Impacts on Students’ Engagement and Performance in High School. New York: Manpower Demonstration Research Corporation. Liebman, Jeffrey B. 1998. “The Impact of the Earned Income Tax Credit on Incentives and the Income Distribution.” In Tax Policy and the Economy, edited by James Poterba. Volume 12. Cambridge, Mass.: MIT Press. Machin, Stephen, and Alan Manning. 1994. “Minimum Wages, Wage Dispersion and Employment: Evidence from the UK Wages Councils.” Industrial and Labor Relations Review 47(2): 319–29. MaCurdy, Thomas. 2004. Evaluating State EITC Options for California. San Francisco: Public Policy Institute of California. Meyer, Bruce D., and Dan T. Rosenbaum. 2001. “Welfare, the Earned Income Tax Credit, and the Labor Supply of Single Mothers.” Quarterly Journal of Economics 116(3): 1063–1114.

76

/

Alternative Labor-Market Policies Neumark, David. 2006. “Evaluating Program Effectiveness: A Case Study of the School-toWork Opportunities Act in California.” Economics of Education Review 25(3): 315–26. ———. 2007. Improving School-to-Work Transitions. New York: Russell Sage Foundation. Neumark, David, and Ann Allen. 2003. “What Do We Know About the Effects of School-toWork? A Case Study of Michigan.” Journal of Vocational Education Research 28(1): 59–84. Neumark, David, and Olena Nizalova. 2007. “Minimum Wage Effects in the Longer Run.” Journal of Human Resources 42(2): 435–52. Neumark, David, and Donna Rothstein. 2006. “School-to-Career Programs and Transitions to Employment and Higher Education.” Economics of Education Review 25(4): 374–93. ———. 2007. “Do School-to-Work Programs Help the “Forgotten Half?” In Improving School-to-Work Transitions, edited by David Neumark. New York: Russell Sage Foundation. Neumark, David, Mark Schweitzer, and William Wascher. 2004. “Minimum Wage Effects Throughout the Wage Distribution.” Journal of Human Resources 39(2): 425–50. ———. 2005. “The Effects of Minimum Wages on the Distribution of Family Incomes.” Journal of Human Resources 40(4): 867–94. Neumark, David, and William Wascher. 1995. “Minimum Wage Effects on Employment and School Enrollment.” Journal of Business and Economics Statistics 13(2): 199–206. ———. 2000. “The Effect of New Jersey’s Minimum Wage Increase on Fast-Food Employment: A Reevaluation Using Payroll Records.” American Economic Review 90(5): 1362–96. ———. 2001a. “Minimum Wages and Training Revisited.” Journal of Labor Economics 19(3): 563–95. ———. 2001b. “Using the EITC to Help Poor Families: New Evidence and a Comparison with the Minimum Wage.” National Tax Journal 54(2): 281–317. ———. 2002. “Do Minimum Wages Fight Poverty?” Economic Inquiry 40(3): 315–33. ———. 2003. “Minimum Wages and Skill Acquisition: Another Look at Schooling Effects.” Economics of Education Review 22(1): 1–10. ———. 2007. “Minimum Wages and Employment.” Foundations and Trends in Microeconomics 3(1–2): 1–182. ———. 2008. Minimum Wages. Cambridge, Mass.: MIT Press. ———. 2009. “Does a Higher Minimum Wage Enhance the Effectiveness of the Earned Income Tax Credit?” Unpublished paper. Irvine: University of California. Orr, Margaret Terry, Thomas Bailey, Katherine L. Hughes, Gregory S. Kienzl, and Melinda Mechur Karp. 2007. “The National Academy Foundation’s Career Academies: Shaping Postsecondary Transitions.” In Improving School-to-Work Transitions, edited by David Neumark. New York: Russell Sage Foundation. Osterman, Paul, Thomas A. Kochan, Richard M. Locke, and Michael J. Piore. 2002. Working in America: A Blueprint for the New Labor Market. Cambridge, Mass.: MIT Press. Person, Ann E., and James E. Rosenbaum. 2007. “Labor-Market Linkages Among Two-Year College Faculty and Their Impact on Student Perceptions, Effort, and College Persistence.” In Improving School-to-Work Transitions, edited by David Neumark. New York: Russell Sage Foundation. Rothstein, Jesse. 2007. “The Unintended Consequences of Encouraging Work: Is the EITC As Good As an NIT?” Unpublished paper. Princeton, N.J.: Princeton University.

/

77

Making the Work-Based Safety Net Work Better Ryan, Paul. 2001. “The School-to-Work Transition: A Cross-National Perspective.” Journal of Economic Literature 39(1): 34–92. Scholz, John Karl. 1994. “The Earned Income Tax Credit: Participation, Compliance, Antipoverty Effectiveness.” National Tax Journal 47(1): 63–87. ———. 2007. “Employment-Based Tax Credits for Low-Skilled Workers.” Hamilton Project Discussion Paper. Washington, D.C.: Brookings Institution. Stern, David. 1984. “School-Based Enterprise and the Quality of Work Experience.” Youth & Society 15(4): 401–27. Stern, David, Neal Finkelstein, James R. Stone, John Latting, and Carolyn Dornsife. 1995. School to Work: Research on Programs in the United States. London and Washington, D.C.: Falmer Press. Stern, David, James Stone, Charles Hopkins, Martin McMillion, and Robert Crain. 1994. School-Based Enterprise: Productive Learning in American High Schools. San Francisco: Jossey-Bass. Stern, David, Christopher Wu, Charles Dayton, and Andrew Maul. 2007. “Learning by Doing Career Academies.” In Improving School-to-Work Transitions, edited by David Neumark. New York: Russell Sage Foundation. Stone, James R., III, and Oscar A. Aliaga. 2007. “Participation in Career and Technical Education and School-to-Work in American High Schools.” In Improving School-to-Work Transitions, edited by David Neumark. New York: Russell Sage Foundation. U.S. General Accounting Office. 1990. Preparing Noncollege Youth for Employment in the U.S. and Foreign Countries. Washington, D.C.: U. S. General Accounting Office. William T. Grant Foundation. 1988. The Forgotten Half: Pathways to Success for America’s Youth and Young Families. Washington, D.C.: Youth and America’s Future, The William T. Grant Commission on Work, Family and Citizenship. Wu, Ximing, Jeffrey M. Perloff, and Amos Golan. 2006a. “Effects of Taxes and other Government Policies on Income Distribution and Welfare.” Unpublished paper. Berkeley: University of California. ———. 2006b. “Effects of Government Policies on Urban and Rural Income Inequality.” \Review of Income and Wealth 52(2): 213–35.

78

/

Chapter 3 On Work and Health Among the American Poor Jayanta Bhattacharya and Peter Richmond Now John Henry, he hammered in the mountains, His hammer was striking fire. But he worked so hard, he broke his own heart. John Henry laid down his hammer and he died, Lord, Lord, John Henry laid down his hammer and he died. —Traditional

T

he relationship between work and health is an uneasy one. For many, work is a fundamental part of their identity; their well-being depends crucially on doing fulfilling work. Unemployment can have devastating psychological and physical impacts. On the other hand, work is often stressful. Also, work may take away time or energy that would otherwise be devoted to health-promoting activities such as exercise. Papers by Christopher J. Ruhm (2000) and others suggest that, at a population level, increases in unemployment are associated with improvements in health. At the same time, losing a job is the source of much stress, which is not healthy. Also, being acutely ill may make it impossible to work in the short run, and being chronically ill may make it difficult to hold down a job in the long run. These divergent influences make the net effect of employment on health difficult to ascertain. The health problems associated with work are perhaps most severe for poorer workers. Work for this group can often require repetitive or dangerous tasks. The well-known Whitehall studies (see Marmot et al. 1991) have extensively documented the fact that workers who are lower down a hierarchy suffer the greatest levels of stress and adverse health consequences.1 So, does working help or hurt the health of poor workers? Having a plausible

/

79

Making the Work-Based Safety Net Work Better

answer to this question is especially important given the postreform emphasis of welfare programs on finding jobs for recipients of government aid to the poor. If work damages the health of poor workers, then there is some justification for welfare agencies to train poor workers to learn to cope with or prevent the damaging effects of work. Of course, the particulars of any policy changes will depend on exactly what health effects work has, which groups of workers are most likely to be affected, and how expensive remedial efforts are. Additional policies may be needed to identify and assist poor people who are not healthy enough to work but who would benefit if their health could be improved. One intriguing point raised by research in this area is that work could be indirectly harmful to health by reducing available nonwork time. Ruhm (2000) proposes this mechanism as an explanation for his finding that higher unemployment rates are correlated with improved health. Working takes time away from alternative health-producing activities such as exercise and preparing healthy meals. It also takes time away from activities that (for most people) are psychologically beneficial, such as spending time with family, friends, and neighbors. Working can thus reduce social connectedness outside of the workplace. This point has important policy implications, to the extent that the data confirm its importance. For instance, reformed zoning laws that place work closer to where welfare recipients live might reduce commuting time. Also, affordable housing programs and housing vouchers for the poor might be structured so that recipients and beneficiaries could afford to live closer to work. Despite the importance of this topic, to date no study has systematically examined the effects of work on the health of the poor in the United States. The work in this chapter is a first attempt to fill this gap in the literature. We use data on poor people from the National Health and Nutrition Examination Survey (NHANES), which is a large, nationally representative data set collected by the Centers for Disease Control (CDC) to monitor the health of the American population. We compare the health status of poor workers and nonworkers on a wide variety of metrics, including serum markers of chronic disease and nutritional deficiency, physical activity levels, and measures of social connectedness.2 Finally, we use recently developed methods in causal inference, described by Azeem M. Shaikh and Edward J. Vytlacil (2005), to bound the causal effect of working on health.

THE LITERATURE ON WORK AND HEALTH Worker self-sufficiency depends greatly on the ability of workers to maintain their ability to work over a long period of time. If the work done by a poor worker harms his physical or psychological health, it may be impossible, despite any economic incentives in return-to-work provisions in welfare programs, to maintain self-sufficiency. Thus, the effects of working on health are a critical yet underaddressed topic of the work-based safety net. In this section we review the literature on the link between unemployment and health. Authors have taken two distinct approaches to identifying the effect of

80

/

On Work and Health

work on health. The microeconomics literature examines how job losses, at an individual level, correlate with health status. This approach is appealing because the mechanism of health harm is clear—job loss is stressful and stress can be harmful to health. The drawback of this approach is that there are almost certainly unmeasured health status events that can lead to both job losses and poor subsequent health, so any measured correlation may not reflect the causal effect of work on health. The macroeconomics literature focuses on how changes in regional unemployment rates relate to changes in population health. The drawback with this approach is that there is no obvious mechanism relating regional unemployment to individual health status. Whereas the microeconomics literature generally finds that loss of employment tends to harm health, the macroeconomic literature tends to find that higher unemployment improves health. We discuss possible mechanisms that may reconcile the findings of these two conflicting literatures. Finally, we review the limited literature on the link between unemployment and health among the poor.

Health and Unemployment A large body of research has found that unemployment is associated with higher mortality risk. Joan K. Morris, Derek G. Cook, and A. Gerald Shaper (1994) document that loss of employment or retirement among British middle-aged men is associated with a nearly two-fold increase in the risk of mortality. Ulf-G. Gerdtham and Magnus Johannesson (2003) find, on the basis of data from a sample of 30,000 Swedes of working age, that unemployment is associated with a 50 percent increase in the age-specific annual mortality rate. The risk of dying from suicide and diseases other than cardiovascular disease and cancer increased significantly. William T. Gallo et al. (2000) also report a positive association between job loss and mortality among older workers. Other work examining health status has linked unemployment with declining health. A study by Tomas Korpi (2001), using a comprehensive measure of physical health, finds significant evidence of worsening health status resulting from unemployment. A potential source of bias in much of this work is that poor health reduces the capacity to work; that is, poor health is a cause of unemployment rather than the other way around (Currie and Madrian 1999). Many of these studies attempt to control for this bias, typically by including observed measures of health status as control variables in their statistical analyses or by comparing the health of individuals before and after loss of employment. Although these approaches go some way in isolating the causal effect of loss of employment on health, they are not entirely satisfactory. In particular, the possibility remains that there are important unobserved events (such as sudden and new health shocks) that lead to both worse subsequent health and to unemployment, even in the absence of any direct causal relationship between work and health. There has been some work on the mechanisms mediating the link between not having a job and health status. A number of studies have examined the effect of

/

81

Making the Work-Based Safety Net Work Better

employment status on alcohol consumption, but the evidence of a relationship is weak and results of such studies are mixed (Muustonen, Paakkanen, and Simpura 1994). One study on a large sample of middle-aged British men by Morris, Cook, and Shaper (1992) indicates that loss of employment is associated with an increased likelihood of gaining weight but is not associated with changes in cigarette and alcohol consumption. These findings suggest that population-level trends in alcohol and tobacco consumption related to the rate of unemployment cannot be explained by the behavioral changes of the individuals experiencing changes in job status. Colin D. Mathers and Deborah J. Schofield (1998) conclude in a review of the literature on health and unemployment that there is reasonably good evidence that unemployment itself is detrimental to health and has an impact on health outcomes: increasing mortality rates and causing physical and mental ill health, as well as greater use of health services. Other analysts have reached the same conclusion (Jin, Shah, and Svoboda 1995; Dooley, Fielding, and Lennart 1996). In fact, this literature tends to find stronger evidence of a negative psychological effect of unemployment than of a negative physical effect (Murphy and Athanasou 1999). In general, these literature reviews have been more accepting than we are of the idea that the microeconomic literature has resolved the causality issues underlying the work-health relationship, given our reading of the papers referenced above (though some of the papers do a better job on the topic than others).

Health and the Unemployment Rate In contrast to the literature just discussed, authors who have looked at the effect of macroeconomic measures (such as the unemployment rate) on health have documented a surprising countercyclical relationship. Here we describe the findings from the main papers in this literature, and we then turn to papers that examine possible mechanisms that could explain this surprising relationship between unemployment and health. Among the hypothesized mechanisms are (1) that unemployment increases leisure time available to people, which in turn leads to healthier living; (2) that the demand for unhealthy activities and risk taking decrease during periods of high unemployment (perhaps because incomes decline); (3) that work itself is inherently stressful or dangerous; and (4) that economic downturns tend to reduce commuting and hence air pollution. Ruhm (2000) uses U.S. mortality data from 1972 to 1991 to document that overall mortality (as well as mortality for eight of ten specific causes) is negatively associated with changes in the unemployment rate.3 He finds that a 1 percent increase in unemployment is associated with a reduction in overall mortality of 0.5 percent. The reduction in mortality is due primarily to a reduction in the probability of developing chronic diseases, rather than fewer accidents. Death from suicide, unsurprisingly, is an exception to the countercyclical trend: suicide rates increase as the unemployment rate increases. Ruhm (2003) also finds that other health outcomes, including heart disease, self-reported

82

/

On Work and Health

medical problems, and restricted-activity days, also follow a counter-cyclical economic pattern. Others describe a similar effect of unemployment on mortality abroad. Eric Neumayer (2004) finds mortality rates rose during economic booms in sixteen German states between 1980 and 2000. Gerdtham and Ruhm (2006) find a similar pattern in twenty-three countries that are members of the Organization for Economic Cooperation and Development between 1960 and 1997. Like the literature on the link between employment status and health status, there has been much work considering mechanisms that link changes in the macroeconomy to health status. Spurts of economic growth might increase work hours and decrease available leisure time, and economic downturns might decrease work hours and increase leisure time. This results in leisure time becoming more valuable during economic upswings, causing health-producing behaviors, such as in-home preparation of meals, physical exercise, and doctors’ visits, to become more costly in terms of leisure time. If economic upturns have the expected effect on leisure time, individuals would likely become more sensitive to the “time price” of food and eat even greater amounts of calorie-rich prepared and processed foods. Ruhm (2005) finds increased obesity during periods of economic expansion, along with less healthful eating habits (2000). Using data from the Behavioral Risk Factor Surveillance Study (BRFSS) for the years 1984 to 1999, Shin-Yi Chou, Michael Grossman, and Henry Saffer (2004) examine the effects of price and availability of fast food and restaurants on obesity. The BRFSS is a large and nationally representative telephone health-tracking survey conducted by the CDC. They find that obesity is negatively related to the cost of fast food and of home-prepared food. Chou, Grossman, and Saffer also find that obesity is negatively correlated with the “time price” (as measured by travel time and waiting time) of obtaining meals outside the home. Ruhm (2005) finds that physical activity decreases during economic upswings, further suggesting a link between the value of leisure time and the regularity of health-producing behaviors. The effect of economic expansion on physical activity may be greater now than in previous decades due to the increasing prevalence of low physical activity jobs identified by Darius Lakdawalla and Tomas Philipson (2002). The more common such jobs are, the more likely it is that economic spurts, which might extend work hours, also result in increased sedentary time. Another consequence of decreased leisure time during periods of economic growth might be that medical care is reduced because the time cost becomes greater. Jessica Primoff Vistnes and Vivian Hamilton (1995) document a decrease in doctor visits for children whose parents work longer hours, suggesting that medical care might also decrease for adults when work hours increase. However, Ruhm (2000, 2003) finds no evidence of this. Health outcomes might also respond to economic fluctuations if products harmful to health, such as alcohol and tobacco, are normal goods. If so, then recessions, which decrease incomes, would also decrease consumption of alcohol and tobacco. Ruhm (2000, 2005) presents strong evidence that smoking decreases dur-

/

83

Making the Work-Based Safety Net Work Better

ing economic downturns, as measured by changes in unemployment rates. Work by Ruhm and William E. Black (2002) finds that alcohol consumption also decreases during recessions. Earlier research comes to similar conclusions (Ruhm 1995; Freeman 1999). The influence of income on risk-taking behavior offers an additional mechanism. In work by Ruhm (1995, 2000) vehicle fatalities, fatal accidents, and homicides are all found to fall when unemployment rates rise. This effect appears to be strongest among younger populations (those twenty to forty-four years old). It is unclear how and why economic activity affects risk-taking behavior, but the correlation appears to be strong. Of course, accidents explain only a fraction of the increased mortality occurring during recessions. The literature has considered nonbehavioral mechanisms as well to explain the negative relationship between the economic cycle and health. Employment can harm health as a result of hazardous working conditions, job-related stress, and the physical and mental taxation of work. Increased labor force participation resulting from economic expansion would likely make the effect of these factors on health more severe. Dean Baker (1985) reviews two types of work stressors affected by economic cycles. Work environment stressors, which include physical hazards such as noise and temperature extremes, chemical hazards, physical danger, physical exertion, and ergonomics, are consistent with hazardous working conditions and are considered more physically than psychologically harmful. Work stressors, which include workload, decisionmaking, and deadlines, are considered more psychologically than physiologically harmful. Robert A. Karasek and Tores Theorell (1990) also document the sources and consequences of jobrelated stress. The degree to which such work stressors can impact health is uncertain, but it seems likely that increased work hours resulting from economic expansion would exacerbate those effects. Research examining the effect of long work hours on the heart health of Japanese men sheds light on possible health consequences. Shigeru Sokejima and Sadanobu Kagamimori (1998) find that for Japanese men, a daily increase in work time of more than three hours increases heart attack risk by about 2.5 times the level of those with a daily work increase of an hour or less. Other work on Japanese male populations has found an association between long work hours and high blood pressure (Hayashi et al. 1996; Iwasaki et al. 1998). Kenneth Y. Chay and Michael Greenstone (2003) document an alternative nonbehavioral mechanism that does not act through increased workload. Their work finds an association between decreased air pollution, which can result from economic recession, and lower infant mortality rates. Although the adult population is generally less sensitive to air pollution, such pollution could have health consequences for certain vulnerable adult populations.

The Impact of Unemployment on the Poor Much evidence supports the notion that people with low incomes are more likely to have poor health outcomes than people with higher incomes. Susan L. Ettner 84

/

On Work and Health

(1996), for instance, measures a positive relationship between permanent income and physical and mental health (as measured by self-assessed health status, work and functional status, bed days, and depressive symptoms). However, even though unemployment tends to lower incomes, this evidence is not dispositive on the direct effect of being out of work on health. There may be a direct effect of employment that this evidence misses. Unfortunately, the issue of health and employment among the poor has been a neglected area of research, though there is some work tangentially related to the topic. A study by Stephen E. Snyder and William N. Evans (2002) examining Social Security beneficiaries suggests that moderate employment could have health benefits for the elderly. They find that mortality was reduced for beneficiaries affected by a change in legislation lowering their Social Security incomes. They attribute the reduction to higher postretirement employment among affected beneficiaries. Snyder and Evans say that “work could have positive health benefits if the work keeps the seniors connected to the community and reduces social isolation” (2003, 2). Though this finding was observed in a very different population, these effects could also apply to poor populations receiving government assistance. Work by J. Blake Turner (1995), however, suggests that unemployment may be less harmful to the health of less well-educated populations than to other populations. Turner analyzes a sample of individuals who were involuntarily terminated from their jobs. He finds that people with a college education and living in an area with low unemployment fare worse in their health than those without a college education or living in an area with high unemployment rates. In particular, the former report higher levels of depression and worse physical health. Turner posits that the greater health effects observed in higher-status victims of job loss may be due to greater damage to their “sense of self,” whereas the stress of job loss among lower-status populations expresses itself more as a financial hardship. This literature does not directly examine the effect of employment on health among the poor. Neither does the available literature, as described above, provide clues as to whether the poor might respond differently to unemployment. Further, the literature reveals a tension between population-wide health effects of changes to the unemployment rate and the inversely moving individual-level health effects of changes in employment status. It is difficult to find an explanation for why the macro and micro literatures differ in their conclusions. Our conjecture is that these two literatures are measuring different mechanisms. The macro literature is picking up the increase in leisure time that comes with economic downturns, while the micro literature is picking up the psychological dislocation caused by involuntary time off. Our hope is that by placing these two mechanisms (time available out of work and direct effects of non-work) in a single coherent context, we can help to clear up this contradiction in the literature, at least for poor populations. Finally, research on the link between work and health among the poor is important for understanding whether a work-based safety net can really promote self-sufficiency over the long term. The work we present in this chapter is a first stab at dealing with these issues. /

85

Making the Work-Based Safety Net Work Better

DATA For our examination of the health effects of work, we rely on the National Health and Nutrition Examination Survey (NHANES). We examine a variety of health, physical activity, and social connectedness measures.

National Health and Nutrition Examination Survey We rely on data from the NHANES III for our primary analyses.4 This national survey was conducted between October 1988 and October 1994 and included nearly 34,000 respondents, age two months and older. The survey oversampled blacks, Mexican Americans, younger children, and older persons to assure adequate representation and includes weights to make the sample nationally representative. We limit the sample we analyze to working-age adults between twentyfive and sixty-five, inclusive. NHANES III collects information in the usual survey domains, including demographics (age, sex, and education), income, and work. The survey also collects substantial health information not normally found in surveys, including data from a physical exam conducted by a doctor, blood and urine tests, physical activity measures, and questions about social connectedness. The NHANES III asks respondents about their income and family composition. Using these two pieces of information, the NHANES III calculates whether each respondent family is above or below the official U.S. poverty line, as defined at the time of the survey. Unlike surveys aimed specifically at ascertaining economic conditions (such as the Current Population Survey), the NHANES III measures income using a single question about total income rather than a series of questions about different sources of income. The former technique generates noisier estimates of income and poverty than the latter. Despite this drawback, the poverty measure provides a useful benchmark of relative deprivation (see Bhattacharya, Currie, and Haider 2004). For this study we select people for inclusion if the ratio of their income to the federally determined poverty line is less than 2. This ratio is known as their poverty-income ratio. We experimented with other definitions of poverty, including moving the poverty-income ratio cutoff down to 1.5, but our qualitative conclusions do not change. In addition to the NHANES III, we use data from later NHANES waves (1999 to 2000, and 2001 to 2002) to track trends in many of these outcomes over this period. The sampling scheme for all of the NHANES surveys required sampling households over multiple years to assure national coverage. Because of this, the NHANES cannot be used to measure nationally representative outcomes in a single year. However, they can be used to measure national outcomes for the periods from 1988 to 1991, 1991 to 1994, 1999 to 2000, and 2001 to 2002. In some cases the questionnaires were changed between the earlier and later waves. We do not track trends when that is the case. We also do not analyze the NHANES for 2003 to 2004 because when this chapter was written the National Center for Health Statistics had not yet released data on the employment status of survey respondents. Finally, for 86

/

On Work and Health

our analysis of the causal effect of working, we use only the NHANES III data because the later NHANES data do not include information on county of residence in the public release version. We need this datum to link our instrumental variables (county-level unemployment and state-level minimum wage) to the NHANES records.

Measures We compare outcomes for workers and nonworkers in three different domains: health, physical activity, and social connectedness. We include the health and physical activity domains because the literature implicated employment as having substantive effects (both deleterious and positive) in both. In some cases, adverse values of the health measures that we analyze would immediately impinge on a person’s ability to function or work. For instance, anemic individuals fatigue easily and may experience difficulty concentrating, with obvious implications on the ability to work. For most of the health measures, though, and for all of the physical activity measures, adverse values would have longer-run consequences. Obesity, for instance, is associated with the incidence later in life of many chronic diseases, such as heart disease and diabetes, that can be expensive to treat. The policy implications of a finding that work has an adverse effect on health outcomes with mainly short-run implications are quite different from a finding that work has an adverse effect on health outcomes with only long-run implications. In the former case, policies that prevent or immediately treat the adverse health condition might be needed to enable workers to stay at the job. The latter case would have implications for a wide range of social programs, including Medicare and Social Security. The social connectedness measures that we analyze serve a dual purpose. First, they are a measure of the extent to which work time impinges on activities outside the workplace. Second, they correlate with concrete activities, such as spending time with family, that many people value. Other authors (see Turcotte 2008, for instance) have used similar measures from other surveys for purposes closely related to ours (though for distinct populations). For instance, if work is found to impinge on social connectedness, then policies that reduce wasted time (such as commuting time) by poor workers might be warranted. Table 3.1 shows the definitions that we use for our health outcomes. We use three anthropometric measures of fatness. The first, obesity, is based on the body mass index (BMI) which is objectively measured during the medical assessment. BMI is weight in grams divided by height in meters squared. Obese individuals are those with a BMI over 30. The second and third measures are the subscapular skinfold and triceps skinfold pinch tests. These tests, which involve measuring skin thickness during a standardized pinch, correlate very well with body fat as a percentage of weight (Cronk and Roche 1982). As a measure of nutritional adequacy, we analyze serum folate. Folate is a B vitamin that is vital for a large number of biochemical processes within the body. Inadequate folate intake can result in macrocytic anemia (a form of low blood /

87

Making the Work-Based Safety Net Work Better TABLE 3.1

/

Health Outcome Definitions

Measure

Definition

Obese High triceps skinfold High subscapular skinfold Low folate Diabetes High cholesterol Low HDL High LDL Anemic (female) Anemic (male)

Body mass index ≥ 30 Triceps skinfold ≥ 25 mm Subscapular skinfold ≥ 27 mm Serum folate less than 7 nmol/L Fasting glucose ≥ 125 mg/dL Serum cholesterol greater than 200 mg/dL Serum HDL less than 35 mg/dL Serum LDL more than 130 mg/dL Hemoglobin < 12 g/dL and hematocrit < 36 percent Hemoglobin < 13 g/dL and hematocrit < 39 percent

Source: Authors’ compilation based on data from Fauci et al. 2008.

count) along with other adverse conditions. During pregnancy, adequate folate consumption helps to prevent congenital anomalies such as neural tube defects. In January 1998, the U.S. government mandated that all flour and uncooked cereal grain be fortified with folate. Since then, folate deficiency prevalence has dropped precipitously, even among the poor in the United States (Ulrich and Potter 2006). For data before 1998, such as the NHANES III, folate deficiency is a good measure of nutritional inadequacy. Diabetes is a serious chronic illness caused by a deficiency in the body’s ability to process glucose. The form of diabetes that is most common among adults, type II diabetes, is strongly associated with obesity. The consequences of severe diabetes are adverse and manifold, including hypertension, heart disease, kidney failure, and blindness, though these consequences can generally be delayed with careful management of glucose levels. Measuring the level of glucose in the blood stream while a patient is fasting is a standard way of screening for the presence of diabetes (and is the method followed in the NHANES). High fasting blood-sugar levels indicate the presence of diabetes. High levels of total cholesterol and low-density lipoprotein (LDL) in the blood are good predictors of subsequent heart disease. These substances promote the formation of atherosclerotic plaques in the coronary arteries, which can lead to ischemia (loss of blood flow to heart muscle tissue or heart attacks). Conversely, high levels of high-density lipoprotein (HDL) tend to inhibit the formation of atherosclerotic plaques. We analyze these measures because they are good markers of cardiovascular health that may be affected by work stress or stress from job loss. Anemia is a clinical condition that is caused by inadequate levels of either red blood cells, which carry oxygen throughout the body, or hemoglobin, a molecule that binds oxygen within red blood cells. There are many causes of anemia, including inadequate nutrition and chronic disease. Among the milder conse-

88

/

On Work and Health

quences of anemia are fatigue, dizziness, and inability to concentrate. More severe consequences include death. In the United States, anemia prevalence is closely correlated with poverty (see Bhattacharya, Currie, and Haider 2004). As we have noted, working may reduce time available to engage in healthy physical activity. The NHANES survey includes an extensive set of questions about the extent of physical activities by respondents. We separately analyze indicators of whether respondents have spent any time in the past thirty days walking, jogging, biking, swimming, weightlifting, or engaging in any other exercise. Finally, we analyze several markers of social connectedness outside the workplace that may also be affected by the time constraints of working. These include indicators for whether respondents have had phone contact with their families at least once per day over the previous week, visited with their family at least once in the previous week, visited a neighbor in the previous month, attended church or other religious service in the previous month, and attended any club or group meetings in the previous month.

A COMPARISON OF WORKERS AND NONWORKERS We start our analysis by estimating simple conditional mean differences between workers and nonworkers on the basis of each of our outcome measures. (Recall that we perform all of our analyses on working-age adults with a poverty-toincome ratio below 2). We condition on a simple set of demographic variables: age, race (encoded as white, black, and other race), sex, income (measured by poverty-to-income ratio), education, and marital (or cohabitation) status. We measure education with a series of mutually exclusive indicator variables for whether the respondent has less than a high school education, a high school degree, or education beyond high school. For our analysis of trends in these outcome measures, we run separate regressions for each outcome for each wave of the NHANES conditional on the same set of demographic covariates just listed. We then use the regressions to predict the levels of the outcome variables separately for workers and nonworkers. For these predictions we set the demographic variables at their mean values in the 1988-to1991 NHANES data. This procedure adjusts our trend estimates so that any changes are not due to changes in the demographic composition of the NHANES samples. Table 3.2 shows descriptive statistics from the NHANES III data, shown separately for workers and nonworkers with a poverty-to-income ratio less than 2. Nonworkers are more likely to be obese, by all three measures, and diabetic. They are also more likely to be anemic. However, the differences between workers and nonworkers in blood lipid levels and folate deficiency are insubstantial. For every category, nonworkers are more likely not to exercise than workers. By some measures of social connectedness, such as family visits, nonworkers appear better connected than workers, whereas by other measures, such as active club or group membership, workers appear better connected. Since these are raw mean differ-

/

89

Making the Work-Based Safety Net Work Better TABLE 3.2

/

Descriptive Statistics

Whole Sample Mean

s.d.

Folate deficiency Diabetes High cholesterol Low HDL High LDL High subscap. fold High triceps fold Anemic Obese

.26 .071 .44 .11 .73 .28 .28 .098 .29

.44 .26 .5 .31 .44 .45 .45 .30 .45

No walking No weightlifting No jogging No biking No swimming No other exercise

.77 .90 .87 .85 .94 .48

Phone family infrequently Visit family infrequently Never visit neighbor Never go to church Never go to clubs

.73 .28 .51 .35 .82

Worker Male Age Poverty-to-income ratio Race: white Race: black Race: Hispanic Race: other Married or cohabiting Education: < high school Education: high school Education: > high school N

.56 .46 43.4 1.1 .19 .37 .39 .053 .54 .56 .29 .15

Nonworkers s.d.

Mean

s.d.

.25 .10 .46 .12 .74 .33 .33 .12 .33

.43 .30 .50 .33 .44 .47 .47 .32 .47

.27 .046 .42 .10 .72 .25 .23 .083 .25

.44 .21 .49 .30 .45 .43 .42 .28 .43

.42 .30 .33 .35 .23 .50

.80 .93 .89 .87 .96 .54

.40 .26 .31 .34 .19 .50

.75 .88 .85 .84 .93 .44

.44 .32 .35 .36 .26 .50

.44 .45 .50 .48 .38

.69 .28 .49 .37 .85

.46 .45 .50 .48 .35

.76 .29 .53 .34 .80

.43 .45 .50 .47 .40

.50 .50 11.3 .52 .39 .48 .49 .23 .50 .50 .46 .35

Mean

Workers

0 .36 46.5 .93 .17 .39 .37 .060 .46 .66 .24 .10

5,085

0 .48 12.2 .51 .38 .49 .48 .24 .50 .47 .43 .30 2,239

1 .53 41.9 1.2 .20 .36 .4 .049 .59 .48 .34 .18

0 .50 10.6 .50 .40 .48 .49 .21 .49 .50 .47 .39 2,845

Source: Authors' compilation based on the NHANES III.

ences, however, they should not be overinterpreted. In particular, there may be many salient differences between workers and nonworkers, which have nothing to do with working per se, that may explain these differences in outcomes. Among the demographic variables alone, workers are more likely to be male, better educated, younger, married, and (even in this comparison of poor people) richer.

90

/

On Work and Health FIGURE 3.1

/

OLS Estimates of Effect of Work on Health Point Estimate

95 Percent Confidence Band

Female

Male

Obese Low HDL

Health Measures

High LDL High Cholesterol High Triceps Skin High Subscapular Skin Folate-Deficient Diabetes Anemia −.1

−.05

0

.05

−.1

−.05

0

.05

Treatment Effect Source: Authors' compilation based on the NHANES III.

Note: The lines on the graph depict regression-adjusted estimates of the effect of working on each outcome. We adjust for race, education, family income, and age. The sample consists of poor individuals with a poverty-to-income ratio less than 2.

The next set of results adjusts for the observed demographic differences (sex, age, poverty-to-income ratio, race, marital status, and education) between workers and nonworkers. Because working is likely to affect the health of men and women differently, we report separate regression results by sex. Figure 3.1 plots the point estimates on the worker dummy and the associated 95 percent confidence intervals for each of the health outcomes. A point estimate to the right of zero implies workers are more likely to be afflicted with some condition, and a point estimate to the left implies workers are less likely. In other words, negative point estimates (left of the bar) denote improvement, and positive point estimates (right of the bar) signify a deleterious effect. Among men the only statistically significant differences are for diabetes and anemia, though workers appear healthier on the basis of every health status measure that we examine. Among women, most of the statistically significant measures suggest that workers are healthier: less obese, better blood lipid readings, and less likely to be diabetic. Figure 3.2 shows worker-to-nonworker differences in physical activity mea-

/

91

Making the Work-Based Safety Net Work Better FIGURE 3.2

/

OLS Estimates of Effect of Work on Physical Activity Point Estimate

95 Percent Confidence Band

Female

Male

No weightlifting

Physical Activity Measures

No swimming

No biking

No jogging

No walking

No other exercises −.15

−.1

−.05

0

.05

−.15

−.1

−.05

0

.05

Treatment Effect Source: Authors' compilation based on the NHANES III.

Note: The lines on the graph depict regression-adjusted estimates of the effect of working on each outcome. We adjust for race, education, family income, and age. The sample consists of poor individuals with a poverty-income ratio less than two.

sures, adjusted for demographic differences. As in figure 3.1, negative point estimates signify improvement and positive point estimates signify worsening. Like the unadjusted numbers in table 3.2, these regressions show that for both male and female workers, workers are more likely than nonworkers to get some exercise, whether it involves walking, running, jogging, biking, weightlifting, swimming, or something else. Not all of these differences are statistically significant, but for both men and women, some of them are. Working apparently does not take so much time away that it prevents poor workers from getting some exercise (relative to the nonworkers).5 Finally, figure 3.3 shows the adjusted differences between workers and nonworkers on our social connectedness measures. For women, the conclusions are similar to the unadjusted numbers: workers are less likely to have frequent contact with their neighbors or family, but are more likely to attend club meetings or church. For men, there is no statistically significant difference between workers

92

/

On Work and Health FIGURE 3.3

/

OLS Estimates of Effect of Work on Social Contacts Point Estimate

95 Percent Confidence Band

Female

Male

Social Contact Measures

No neighbor visit

Infrequent family visit

Infrequent family phone

No clubs

No church −.1

−.05

0

.05

.1

−.1

−.05

0

.05

.1

Treatment Effect Source: Authors' compilation based on the NHANES III. Note: The lines on the graph depict regression-adjusted estimates of the effect of working on each outcome. We adjust for race, education, family income, and age. The sample consists of poor individuals with a poverty-to-income ratio less than 2.

and nonworkers in their contact with family or neighbors. Like the women, male workers are more likely than male nonworkers to attend club or group meetings. They are also more likely to attend church services. What do all these results mean regarding the self-sufficiency of workers? If we interpret these differences as causal, they suggest that working improves the physical health of both male and female workers on several clinically important dimensions. They also suggest that working leads to more time spent on physical activity and (for male workers) on social activities outside of work. If this interpretation is right, programs that promote work by the poor lead to a virtuous cycle. They cause poor people to invest more in their health, in their family, and in their community, which in turn promotes long-run self-sufficiency, enabling these people to stay at work. However, this pattern of results has an alternative explanation. Perhaps the working and nonworking poor differ on bases other than their work behavior that we as analysts do not observe. The working poor may be

/

93

Making the Work-Based Safety Net Work Better

more efficient in their use of time for the production of health and social connectedness. Thus, we need estimates of the effects of working that can plausibly be interpreted as causal.

TRENDS, 1988 TO 2002 Since the NHANES III was collected between 1988 and 1994, a number of important changes have occurred that may potentially alter the effect of working on health. Among these are the continuing rise in obesity prevalence, the development and expanding use of statin medications to control adverse blood lipid levels, and the federal mandate to fortify flour with folate. In this section we analyze data from later waves of the NHANES (1999 to 2000 and 2001 to 2002) to see how health and physical activity outcomes have evolved since 1994. We do not analyze trends in social connectedness outcomes because the later NHANES waves do not include those questions in the same form as they were included in the NHANES III. These trend data are adjusted for the same set of demographic covariates that we used in the previous section. The trends suggest the possibility that there may be some noncontroversial policy interventions and technological innovations that break the link between work and health. Figure 3.4 shows how the adjusted body mass index and body fat measures have changed between 1988 and 2002 for male and female workers and nonworkers. BMI trends show substantial increases for all groups, confirming the existence of an obesity epidemic. The increases for workers and nonworkers and for each sex overlap—the rise has been no greater for workers than for nonworkers. By contrast, subscapular skinfold measures and triceps skinfold measures have remained unchanged, and in the case of subscapular skinfold trends for women have actually declined. This suggests that not all of the rise in body weight among the poor in the United States has been due to increased body fat. Figure 3.5 shows how our adjusted measures of serum lipids have changed for workers and nonworkers (again separately for men and women). One might expect that with the increasing use of statins, there would be a decline in the prevalence of hypercholesterolemia (high cholesterol) or high LDL levels, especially among workers who would most likely be able to afford the medication. However, this is not the case. In fact, hypercholesterolemia prevalence has increased among male workers and declined among female nonworkers. The prevalence of high LDL levels has declined among female workers and nonworkers, as well as for male nonworkers; it has remained flat for male workers. Finally, the prevalence of low levels of HDL have remained flat for female workers and nonworkers and increased for male workers and nonworkers. That there has not been an improvement in HDL levels for any group is not particularly surprising, since the most commonly used statins during the observed period did not have any direct effect on HDL levels. Figure 3.6 shows trends in the prevalence of diabetes, folate deficiency, and anemia. Consistent with the rise in BMI observed in figure 3.4, diabetes prevalence

94

/

On Work and Health FIGURE 3.4

/

Trends in Anthropometric Measures (Adjusted for Demographics) Body Mass Index Trends Male

Female

BMI

32 30 28 26 1990

1995

2000

2005

1990

1995

2000

2005

2000

2005

2000

2005

Subscapular Skinfold Trends Male

Female

mm

25

20

15 1990

1995

2000

2005

1990

1995

Triceps Skinfold Trends Male

Female

mm

25 20 15 10 1990

1995

2000

2005

Workers

1990

1995 Nonworkers

Source: Authors' compilation based on the NHANES III, NHANES 1999 to 2000, and NHANES 2001 to 2002. Note: The sample consists of poor individuals with a poverty-to-income ratio less than 2.

/

95

Making the Work-Based Safety Net Work Better FIGURE 3.5

/

Trends in Lipid Measures (Adjusted for Demographics) P(High Cholesterol) Male

Female

Probability

.55 .5 .45 .4 1990

1995

2000

2005

1990

1995

2000

2005

2000

2005

2000

2005

P(Low HDL Level) Male

Female

Probability

.25 .2 .15 .1 .05 1990

1995

2000

2005

1990

1995

P(High LDL Level) Male

Probability

Female .75 .7 .65 .6 1990

1995

2000

2005

Workers

1990

1995 Nonworkers

Source: Authors' compilation based on the NHANES III, NHANES 1999 to 2000, and NHANES 2001 to 2002. Note: The sample consists of poor individuals with a poverty-to-income ratio less than 2.

96

/

On Work and Health FIGURE 3.6

/

Other Health Trends (Adjusted for Demographics) P(Folate-Deficient) Male

Female

Probability

.3 .2 .1 0 1990

1995

2000

2005

1990

1995

2000

2005

2000

2005

2000

2005

P(Anemic) Male

Probability

Female .15 .1 .05 0 1990

1995

2000

2005

1990

1995

P(Diabetic)

Probability

Female

Male

.15 .1 .05 1990

1995

2000

2005

Workers

1990

1995

Nonworkers

Source: Authors' compilation based on the NHANES III, NHANES 1999 to 2000, and NHANES 2001 to 2002. Note: The sample consists of poor individuals with a poverty-to-income ratio less than 2. /

97

Making the Work-Based Safety Net Work Better

has risen in parallel for workers and nonworkers. Folate deficiency rates have fallen for all groups, which suggests that the fortification mandates imposed in 1998 have successfully reduced folate deficiency. Finally, the prevalence of anemia has fallen mildly for all groups except male workers, for whom it has remained flat. Figure 3.7 shows select physical activity trends. Among workers and nonworkers and for males and females, there have been increases in the frequency of walking, jogging, and weightlifting. Because there are many disparate results, we provide table 3.3, which summarizes the direction of each of the trends for each group: male and female workers and nonworkers. The most striking thing about this table is that for nearly every measure we examine, the trends for workers and nonworkers move in the same direction. Apparently, whatever forces shape these trends affect workers and nonworkers in similar ways. Another lesson to be drawn from analyzing these trends is that simple policy innovations and technological change in medicine can have a profound effect on the relationship between work and health. Prior to the folate supplementation of the food supply, for instance, poor female workers were more likely to be folatedeficient than poor female nonworkers (though the result was not statistically significant). Mandated supplementation drove the rate of folate deficiency in this population to nearly zero, breaking any possible link between work and this particular aspect of health. A lack of health insurance coverage may be an important reason for the gap between poor workers and nonworkers in health outcomes. Workers often lose their eligibility for Medicaid after starting work. If low-wage jobs do not provide health insurance (and they often do not), then poor workers may face challenges in availing themselves of breakthrough medical technologies. For instance, the new generation of statin medications hold the promise of raising serum HDL (“good” cholesterol) levels, while reducing serum LDL (“bad” cholesterol) levels. Policies that encourage the coverage of the newer statin medications by Medicaid and private health insurance might reduce the gap between workers and nonworkers in the likelihood of adverse serum lipid profiles. Policies that extend Medicaid to poor workers who were formerly on Medicaid and who do not have insurance through their workplace might reduce gaps for a wide range of health outcomes.

IDENTIFYING THE TREATMENT EFFECT OF WORKING To identify the causal effect of working on health status, we need an instrumental variable—something that is strongly correlated with whether an individual is working, but that is arguably uncorrelated with the unobserved determinants of the health, physical activity, and social connectedness outcomes. In an important sense, an instrumental variable is a randomizing variable that serves a role similar to that of the simple coin flip in a randomized controlled trial. The two key fea-

98

/

On Work and Health FIGURE 3.7

/

Trends in Physical Activity Measures (Adjusted for Demographics) Walking or Biking in Past Thirty Days Male

Female

Times per Month

10

5

0 1990

1995

2000

2005

1990

1995

2000

2005

2000

2005

Weightlifting in Past Thirty Days Male

Female 4

Times per Month

3

2

1

0 1990

1995

2000

2005

Workers

1990

1995

Nonworkers

Source: Authors' compilation based on the NHANES III, NHANES 1999 to 2000, and NHANES 2001 to 2002. Note: The sample consists of poor individuals with a poverty-to-income ratio less than 2. /

99

Making the Work-Based Safety Net Work Better TABLE 3.3

/

Summary of Results

Women

Body mass index Subscapular skinfold Triceps skinfold Diabetes High cholesterol Low HDL High LDL Anemia Folate deficiency Walk or bike frequency Weightlifting

Men

Workers

Nonworkers

Workers

Nonworkers

up down flat up up flat down down down

up down flat up flat or down flat down down down

up flat up (a bit) up up up down flat down

up flat up (a bit) up flat up down down down

up up

up up

flat or up up

up up

Source: Authors' compilation based on the NHANES, various years.

tures of the coin flip are that it assigns treatment and that it has nothing to do with outcomes in the different branches of the trial once treatment is assigned. These requirements for a coin flip are directly analogous to the requirements of a good instrumental variable. Here we consider two different instrumental variables: the unemployment rate in the county and the minimum wage in the state where each NHANES III resident lives.6 In the labor economics literature, the local unemployment rate has often been shown as a powerful correlate of whether an individual is working (see, for instance, Black, McKinnish, and Sanders 2003). This is true in our data as well. We calculate the average value of the monthly unemployment rate within each county for 1988 to 1994, inclusive (mean = 6.23 percent; median = 5.73 percent; standard deviation = 1.85 percent). For workers, the mean local unemployment rate in this period was 6.13 percent and for nonworkers, it was 6.35 percent. Let z1 be an indicator of whether the local unemployment rate is above its mean value of 6.23 percent. In our data, P[working |z1 = 1] = 0.529 and P[working |z1 = 0] = 0.552, so, clearly, living in a county with high local unemployment is correlated with not working. It is, of course, impossible to prove that z1 is uncorrelated with the unobserved determinants of health, physical activity, or social connectedness. However, there is at least a prima facie case to be made that it is. The local unemployment rate is a reflection of the economic vigor of a community, which is determined largely by the macroeconomic forces affecting that community. Given, individuals, especially poor individuals, have little control over the unemployment rate in the counties where they live. Of course, people may choose where they live in part on the basis of the local unemployment rate. However, to the extent that the transac-

100

/

On Work and Health

tion costs of moving are high, people will take where they live as a given. Under such assumptions, local unemployment meets the criteria of a good instrument. The case against local unemployment rate as an instrument comes from the literature that macroeconomic cycles are directly related to health status. However, the mechanisms considered in that literature, such as the stress of working, run indirectly through the effects of working rather than through the cycles themselves. Indeed, it is difficult to think of any direct mechanism that links macroeconomic data to the health status of any given individual. This is true for the other outcomes we consider as well. So to us this does not seem like a good argument against using the local unemployment rate as an instrument. Nevertheless, it is prudent to consider other instruments to test the robustness of our findings. As a second instrument, we consider the minimum wage in the state where each individual lives.7 This is not the appropriate venue to consider whether the minimum wage is negatively correlated with employment in the economy at large; as David Neumark (chapter 2, this volume) and the literature discuss, this has been a very contentious issue. We will simply note here that in our sample, the minimum wage is negatively correlated with whether an individual is working. The mean value of the minimum wage variable in our sample is $3.16 per hour, with a standard deviation of $1.11. For workers, the mean value is $3.12 (standard deviation $1.14) and for nonworkers it is $3.22 (standard deviation $1.05). Let z2 be an indicator of whether the minimum wage is above its mean value in our sample. For our NHANES respondents, P[working |z2 = 1] = 0.667, while P[working |z1 = 0] = 0.722. Our sample consists of poor people with (typically) low levels of skills and education. Standard economic theory predicts that this is a group in which we should see the greatest negative employment effect of the minimum wage, and this is exactly what we observe. Our argument for whether the minimum wage is uncorrelated with the unobserved determinants of health, physical activity, and social connectedness is directly analogous to our argument that local unemployment is a good instrument. People generally, and the people in our sample in particular, have no special control over how minimum wages are set in their state—they take it as a given. To the extent that the costs of moving are relatively high, people take their location (and hence the minimum wage) as fixed. This assumption is even more plausible than in the case of the local unemployment rate, since the costs of moving to a new state are presumably higher than the costs of moving to a new county within a state.

CAUSAL EFFECT ESTIMATES In this section we present bounds on the causal effect of working on health using the instruments (quasi-randomizing variables) that we justify in the previous section. We briefly describe the method we use to generate these bounds—known as Shaikh-Vytlacil (SV) bounds—in the appendix to this chapter. We derive treatment effect estimates separately for each of our instruments, z1 (local unemploy-

/

101

Making the Work-Based Safety Net Work Better FIGURE 3.8

/

SV Bounds: Effect of Work on Health (Females) IV: County Unemployment Rate

IV: State Minimum Wage

Obese Low HDL

Health Measures

High LDL High cholesterol High triceps skin High subscapular skin Folate-deficient Diabetes Anemia −.5

0

.5

−.5

0

.5

Treatment Effect Bound Source: Authors' compilation based on the NHANES III.

ment rate) and z2 (minimum wages). Our strategy is to conclude that working has a causal effect in a given direction only if both instruments imply SV bounds on the treatment effect of working in the same direction. If the instruments point in different directions, we interpret the data as inconclusive on the direction of the causal effect. Figures 3.8 and 3.9 show the SV bounds for the health outcomes for females and males. All of the figures in this section are configured so that a bar to the left of zero indicates that working decreases the probability of the outcome, and a bar to the right of zero indicates that working increases the probability of the outcome. All of the outcomes are nonpreferred, meaning that negative bounds signify an improvement and positive bounds signify worsening. For females, working increases the probability of high cholesterol, high LDL levels, and high subscapular skinfold measurements. It decreases the probability of obesity and folate deficiency, and it has inconclusive effects on the other health outcomes. For males, working increases the probability of high cholesterol, folate deficiency, and diabetes. It decreases the probability of low HDL levels and anemia. For both men and women, the measured causal effects of working are mixed; from this pattern

102

/

On Work and Health FIGURE 3.9

/

SV Bounds: Effect of Work on Health (Males) IV: County Unemployment Rate

IV: State Minimum Wage

Obese Low HDL

Health Measures

High LDL High cholesterol High triceps skin High subscapular skin Folate-deficient Diabetes Anemia −.5

0

.5

−.5

0

.5

Treatment Effect Bound Source: Authors' compilation based on the NHANES III.

of results, we cannot say that working either improves or harms the physical health of the poor. Figures 3.10 and 3.11 show the SV bounds on the causal effect of working on physical activity levels. For females, working increases the probability of not biking, jogging, or walking. For males, working increases the probability of not weightlifting, biking, walking, and other exercises. This pattern of results seems like solid evidence that for both sexes working reduces the time available to exercise. Finally, figures 3.12 and 3.13 show the SV bounds on the causal effect of working on social connectedness measures. For females, working increases the probability of infrequent family visits, of infrequent phone contact with family members, and of not attending any club or group meetings. For males, working increases the probability of not visiting neighbors, of infrequent family visits, and of not attending any club or group meetings. As is the case for exercise, the evidence shown here strongly suggests that working reduces the time available to engage in social activities outside of work, even with family members.

/

103

Making the Work-Based Safety Net Work Better FIGURE 3.10

/

SV Bounds: Effect of Work on Physical Activity (Females) IV: County Unemployment Rate

IV: State Minimum Wage

No weightlifting

Physical Activity Measures

No swimming

No biking

No jogging

No walking

No other exercise −.5

0

.5

−.5

0

.5

Treatment Effect Bound Source: Authors' compilation based on the NHANES III.

CONCLUSION In this chapter we have considered the complex relationship between work and health among the poor in the United States. Using evidence from the NHANES, we have compared workers and nonworkers on the basis of several important objective measures of health, as well as on the basis of exercise habits and social connectedness. A comparison of mean outcomes shows that both male and female workers are healthier and more likely to exercise than nonworkers. By some measures, workers have more social contacts out of the workplace than nonworkers, though by other measures the opposite is true. These conclusions remain true even after adjustment for demographic differences between workers and nonworkers. Since the NHANES III finished collection in 1994, the outcome measures we examined have not remained static. The obesity epidemic has continued unabated, statin drugs have been introduced to the market and become widely used, and

104

/

On Work and Health FIGURE 3.11

/

SV Bounds: Effect of Work on Physical Activity (Males) IV: County Unemployment Rate

IV: State Minimum Wage

Physical Activity Measures

No weightlifting

No swimming

No biking

No jogging

No walking

No other exercise −.5

0

.5

−.5

0

.5

Treatment Effect Bound Source: Authors' compilation based on the NHANES III.

the government has required the supplementation of folate levels. Certainly other events that affect the health of the poor but that we do not consider have taken place. We find that for the health trends that we do examine, workers and nonworkers have moved in the same direction. Apparently, the forces that shape these trends do not distinguish between workers and nonworkers.8 We calculated bounds on the treatment effect of working on health, exercise, and social connectedness. Although the mean differences between workers and nonworkers in these outcomes are interesting, they should not be interpreted as measures of the causal effect of working. The mean differences reflect a complex set of decisions, circumstances, and constraints that lead to these outcomes. Our causal estimates suggest that working has a complex set of effects on the health of workers, but it unambiguously decreases the time workers spend on exercising and on social relations with family, neighbors, and friends. To some degree this is inevitable—there are only twenty-four hours in a day, and time spent on one activity cannot be spent on another activity. Diminished leisure time is simply an unavoidable consequence of employment. There may be policy interventions that would lower the cost of health-promot-

/

105

Making the Work-Based Safety Net Work Better FIGURE 3.12

/

SV Bounds: Effect of Work on Social Connectedness (Females) IV: County Unemployment Rate

IV: State Minimum Wage

Social Contact Measures

No neighbor visit

Infrequent family visit

Infrequent family phone

No clubs

No church −.5

0

.5

−.5

0

.5

Treatment Effect Bound

Source: Authors' compilation based on the NHANES III.

ing activities by relaxing time constraints for poor workers. Such policies might include improved public transportation and greater mixed zoning to reduce commuting times; subsidized day care to lower the cost of leisure time for parents; entrepreneurship incentives that would enable workers to create fulfilling jobs; and expanding workplace policies that permit flexible working schedules. Calculating whether such interventions are worth the money is beyond the scope of this paper. Such calculations, when done, should take the benefits we measure for poor workers into account. The main place where we find mixed results is on the causal effect of working on physical health. Both the conditional mean differences (reported in “A Comparison of Workers and Nonworkers”) and the causal effect estimates on physical activity and on social connectedness (reported in “Identifying the Treatment Effect of Working”) point in a consistent direction. We believe that our estimates of the causal effect of working on physical health are mixed because, in truth, the relationship between work and health is complicated and conflicted. For instance, some jobs taken by poor people are physically dangerous. Others are often quite stressful (and it is well known—see the Whitehall

106

/

On Work and Health FIGURE 3.13

/

SV Bounds: Effect of Work on Social Connectedness (Males) IV: County Unemployment Rate

IV: State Minimum Wage

Social Contact Measures

No neighbor visit

Infrequent family visit

Infrequent family phone

No clubs

No church −.5

0

.5

1

−.5

0

1

.5

Treatment Effect Bound

Source: Authors' compilation based on the NHANES III.

studies—that stress harms physical health) (Marmot 2003). These considerations might lead one to conclude that work is bad for your health. On the other hand, work is a fundamental part of identity for many. In stage 7 (the midadult years) of Erik H. Erikson’s (1994) theory of psychosocial development, meaningful work is necessary to avoid mental stagnation, which surely cannot be good for your health. Though our findings provide a valuable first step toward estimating the effect of working on the health of the poor, more work will be necessary to fully explore this relationship and make the results useful in a policy context. Of particular importance is the need to define the scope of the effect of working on health-producing activities. Beyond exercise, are other health behaviors, such as eating habits and regularity of medical care, affected as well? The answer to these questions will determine the magnitude of the effect of working on the health of the poor and guide the development of targeted policies. Further, such research would either support or refute the causal mechanism suggested here, that employment reduces health-producing behaviors because of diminished leisure time. If other health behaviors are found to be sensitive to the availability of leisure time, the

/

107

Making the Work-Based Safety Net Work Better

case for policies that minimize time constraints for the working poor will be strengthened. If not, other, more direct, policies should be explored. Alternatively, research examining the sensitivity of time-consuming health behaviors to varying levels of employment demand (twenty to forty hours per week), could both test our proposed causal mechanism and suggest an additional policy option (namely, promoting part-time work). By linking work hours to health behaviors, such research could also explicate the negative relationship observed between macroeconomic conditions and population health. Finally, an important challenge for future research in this area will be to better understand how health insurance coverage mediates the relationship between working and health. Because many Americans receive their health insurance through their employers, the loss of employment is often accompanied by loss of health insurance.9 If the loss of job occurs because of poor health, the concomitant loss of insurance may cause health to suffer further. On the other hand, the poor can often gain access to health insurance from their state through Medicaid, but may become ineligible if working raises income above Medicaid qualification thresholds. Understanding the role of health insurance will thus require careful economic analysis.

APPENDIX: SHAIKH-VYTLACIL BOUNDS For our analysis of the causal effects of working on the various outcomes, we rely heavily on a nonparametric bounding method developed by Azeem M. Shaikh and Edward J. Vytlacil (2005). In this section we provide a brief description of these methods.10 Recovering causal effects from observational data is difficult. Unlike a randomized trial, the treated population (workers, in our case) differ in systematic ways from the untreated population (nonworkers). In a randomized trial, coin flips or other randomizing technology are used to assign treatment, so that the treated and untreated are balanced with regard to the unobserved determinants of the outcomes. The basic idea of the Shaikh-Vytlacil method is to use an instrumental variable to identify the direction of the treatment effect. The instrumental variable serves as a quasi–coin flip: it must be correlated with whether the worker receives treatment but must not be with the unobserved determinants of the outcomes. In our study, we use local unemployment rates and state minimum-wage laws as instrumental variables. Unlike traditional methods of causal inference in observational data, such as the Heckman selection model, however, the Shaikh-Vytlacil method does not rely on unverifiable functional form assumptions about the unobserved determinants of the outcomes. There are other nonparametric bounding methods available in the literature, such as the method developed by Charles Manski and John Pepper (2000). Unlike the Shaikh-Vytlacil method, these other methods require knowing the direction of the causal effect prior to constructing the estimates, which will not do for our application (see Bhattacharya, Shaikh, and Vytlacil 2008 for details).

108

/

On Work and Health

The Shaikh-Vytlacil bounds require that all of the outcome variables be indicators, which they are. Let Y be an indicator of some outcome and let D be an indicator for whether an NHANES respondent is working. These bounds also require a binary instrument for treatment, Z. To simplify the notation, suppose that Z is ordered so that Pr{Y = 1 | Z = 1} > Pr{Y = 1 | Z = 0}. Consider the following triangular system of equations: Y = r(D,ε) D = s(Z,ν)

3.1

Let Y1 denote the outcome that would be observed if the respondent works and let Y0 denote the outcome that would be observed if the respondent does not work. These potential outcomes are given by Y1=r(1,ε) and Y0=r(0, ε). In this notation, the effect of working on the outcome is Y1 – Y0. The average effect of working on the outcome is therefore E[Y1 – Y0] = Pr{Y1 = 1} – Pr{ Y0 = 1}. We maintain the assumption that (Y,D) is determined by equation 3.1. Furthermore, we assume that either r(1,ε) ≥ r(0,ε) or r(1,ε) ≤ r(0,ε) holds for almost every value of ε. These assumptions guarantee that the treatment varies (weakly) monotonically with the outcome. Finally, we assume that s(1,ν) ≥ s(0,ν) for almost every ν or s(1,ν) ≤ s(0,ν) for almost every ν. This assumption guarantees that the outcome is monotone in treatment and that the treatment is monotone in the instrument, but does not impose the direction of the monotonicity in either case. However, we do not assume any particular functional form on ν or on ε. Under these assumptions, it is impossible to identify the treatment effect of working on the outcomes. However, it is possible to derive bounds on those treatment effects. These bounds impose more structure than the entirely nonparametric Manski bounds. In particular, these bounds require monotonicity, while the Manski bounds do not. However, they do not impose so much structure as to guarantee point identification. The Shaikh-Vytlacil bounds have a form that depends on the sign of Pr{Y = 1 | Z = 1} − Pr{Y = 1 | Z = 0}. Recall that we have ordered Z such that Pr{Y = 1 | Z = 1} > Pr{Y = 1 | Z = 0}. If Pr{Y = 1 | Z = 1} − Pr{Y = 1 | Z = 0} > 0, then the implied bounds (with endpoints BLSV and BUSV) are given by: L = Pr{Y = 1 | Z = 1} − Pr{Y = 1 | Z = 0} BSV U = Pr{D = 1,Y = 1 | Z = 1} + Pr{D = 0 | Z = 1} – Pr{D = 0, Y = 1 | Z = 0}. BSV

If, on the other hand, Pr{Y = 1 | Z = 1} − Pr{Y = 1 | Z = 0} < 0, then the implied bounds are given by: BLSV = Pr{D = 1,Y = 1 | Z = 1} − Pr{D = 0,Y = 1 | Z = 0} – Pr{D = 1 | Z = 0}, BU SV = Pr{Y = 1 | Z = 1} − Pr{Y = 1 | Z = 0};

Finally, if Pr{Y = 1 | Z = 1} − Pr{Y = 1 | Z = 0} = 0, then E[Y1 – Y0] = 0.

/

109

Making the Work-Based Safety Net Work Better

NOTES 1. The Whitehall studies focused solely on British civil servants; it is not clear the extent to which its conclusions extend beyond this select group. 2. Here, “serum markers” means measures of chemicals in the bloodstreams of survey respondents. Many of these chemicals are known to be good at indicating the presence or absence of chronic diseases such as diabetes or atherosclerosis. Later in the chapter, we describe the clinical importance of each marker that we consider. 3. Early work by M. Harvey Brenner (1971, 1973, 1979) suggests a positive association of economic cycles with heart disease mortality, and fetal, infant, and maternal mortality. However, this research has been widely discredited by Adam Wagstaff (1985), among others, because of serious technical flaws. 4. The NHANES are a series of surveys conducted on a periodic basis by the Centers for Disease Control aimed at measuring the health of the U.S. population. The first NHANES survey (NHANES I) was conducted between 1971 and 1975; the second NHANES (NHANES II) was conducted between 1976 and 1980; the third NHANES between 1988 and 1994. In 1999, a new wave of the NHANES was started in two year intervals (NHANES 1999 to 2000; NHANES 2001 to 2002; NHANES 2003 to 2004; and NHANES 2005 to 2006). 5. It is, of course, possible that the nonworkers are disabled and physically incapable of exercising. However, just 9.7 percent of nonworkers and 3.9 percent of workers report any difficulty in walking a quarter of a mile. This difference in disability rates between workers and nonworkers is too small to explain the difference in exercise (walking) rates, especially for male workers. 6. The data on local unemployment rates and on minimum wage ordinances come from the Bureau of Labor Statistics and the Department of Labor. 7. For our instrument, we take the average value of this quantity over the 1988-to-1994 period. 8. Another important limitation of the NHANES data is that it samples the noninstitutionalized population. It thus misses the population of people in jail, which consists predominantly of men. 9. Continuing coverage after separating from an employer through COBRA coverage is typically expensive, though perhaps not as potentially expensive as not having health insurance. 10. For a full development of these bounds, please see Azeem Shaikh and Edward Vytlacil (2005) or Jayanta Bhattacharya, Shaikh, and Vytlacil (2005).

REFERENCES Baker, Dean. B. 1985. “The Study of Stress at Work.” Annual Review of Public Health 6: 367–81. Bhattacharya, Jayanta, Janet Currie, and Steven Haider. 2004. “Poverty, Food Insecurity, and Nutritional outcomes in Children and adults.” Journal of Health Economics 23(4): 839–62. 110

/

On Work and Health Bhattacharya, Jayanta, Azeem Shaikh, and Edward Vytlacil. 2005. “Treatment Effect Bounds: An Application to Swan-Ganz Catheterization.” NBER Working Paper No. 11263. Cambridge, Mass.: National Bureau of Economic Research. ———. 2008. “Treatment Effect Bounds Under Monotonicity Assumptions: An Application to Swan-Ganz Catheterization.” American Economic Review (Papers and Proceedings) 98(2): 351–56. Black, Dan A., Terra G. McKinnish, and Seth G. Sanders. 2003. “Does the Availability of High-Wage Jobs for Low-Skilled Men Affect Welfare Expenditures? Evidence from Shocks to the Steel and Coal Industries.” Journal of Public Economics 87(9–10): 1921–42. Brenner, M. Harvey. 1971. “Economic Changes and Heart Disease Mortality.” American Journal of Public Health 61(3): 606–11. ———. 1973. “Fetal, Infant, and Maternal Mortality During Periods of Economic Instability.” International Journal of Health Services 3(2): 145–59. ———. 1979. “Mortality and the National Economy: A Review, and the Experience of England and Wales, 1936–76.” Lancet 2(8142): 568–73. Chay, Kenneth Y., and Michael Greenstone. 2003. “The Impact of Air Pollution on Infant Mortality: Evidence from Geographic Variation in Pollution Shocks Induced by a Recession.” Quarterly Journal of Economics 118(3): 1121–67. Chou, Shin-Yi, Michael Grossman, and Henry Saffer. 2004. “An Economic Analysis of Adult Obesity: Results from the Behavioral Risk Factor Surveillance System.” Journal of Health Economics 23(3): 565–87. Crimmins, Eileen, Sandra L. Reynolds, and Yasuhiko Saito. 1999. “Trends in Health and Ability to Work Among the Older Working-Age Population.” Journals of Gerontology, Series B: Psychological Sciences and Social Sciences 54(1): S31–S40. Cronk, Christine E., and Alex F. Roche. 1982. “Race- and Sex-Specific Reference Data for Triceps and Subscapular Skinfolds and Weight/Stature.” American Journal of Clinical Nutrition 35(2): 347–54. Currie, Janet, and Briggette C. Madrian. 1999. “Health, Health Insurance and the Labor Market.” In Handbook of Labor Economics, vol. 3, 1st ed., edited by Orley Ashenfelter and David Card. Amsterdam: North-Holland. Dooley, David, Jonathan Fielding, and Lennart Levi. 1996. “Health and Unemployment.” Annual Review of Public Health 17: 449–65. Elinson, Lynn, Patricia Houck, Steven C. Marcus, and Harold Alan Pincus. 2004. “Depression and the Ability to Work,” Psychiatric Services 55(1): 29–34. Erikson, Erik H. 1994. Identity and the Life Cycle. New York: Norton. Ettner, Susan L. 1996. “New Evidence on the Relationship Between Income and Health.” Journal of Health Economics 15(1): 67–85. Fauci, Anthony S., Eugene Braunwald, Dennis L. Kasper, Stephen L. Hauser, Dan L. Longo, J. Larry Jameson, and Joseph Loscalzo, eds. 2008. Harrison’s Principles of Internal Medicine, 18th ed. New York: McGraw-Hill Professional. Freeman, Donald G. 1999. “A note on ‘Economic Conditions and Alcohol Problems.’” Journal of Health Economics 18(5): 661–70. Gallo, William T., Elizabeth H. Bradley, Michele Siegel, and Stanislav V. Kasl. 2000. “Health Effects of Involuntary Job Loss Among Older Workers: Findings from the Health and Retirement Survey.” Journals of Gerontology, Series B: Psychological Sciences and Social Sciences 55(3): 131–40. /

111

Making the Work-Based Safety Net Work Better Gerdtham, Ulf-G. and Magnus Johannesson. 2003. “A Note on the Effect of Unemployment on Mortality.” Journal of Health Economics 22(3): 505–18. Gerdtham, Ulf-G., and Christopher J. Ruhm. 2006. “Deaths Rise in Good Economic Times: Evidence from the OECD.” Economics of Human Biology 4(3): 298–316. Hayashi, Takeshi, Yasuki Kobayashi, Kazue Yamaoka, and Eiji Yano. 1996. “Effect of Overtime Work on 24-hour Ambulatory Blood Pressure.” Journal of Occupational and Environmental Medicine 38(10): 1007–11. Iwasaki, Kenji, Takeshi Sasaki, Tatsuo Oka, and Naomi Hisanaga. 1998. “Effect of Working Hours on Biological Functions Related to Cardiovascular System Among Salesmen in a Machinery Manufacturing Company.” Industrial Health 36(4): 361–67. Jin, Robert L., Chandrakant P. Shah, and Tomislav J. Svoboda. 1995. “The Impact of Unemployment on Health: A Review of the Evidence.” Canadian Medical Association Journal 153(5): 529–40. Karasek, Robert A., and Tores Theorell. 1990. Healthy Work: Stress, Productivity, and the Reconstruction of Working Life. New York: Basic Books. Korpi, Tomas. 2001. “Accumulating Disadvantage: Longitudinal Analyses of Unemployment and Physical Health in Representative Samples of the Swedish Population.” European Sociological Review 17(3): 255–73. Lakdawalla, Darius, and Tomas Philipson. 2002. “The Growth of Obesity and Technological Change: A Theoretical and Empirical Examination.” NBER Working Paper No. 4863. Cambridge, Mass.: National Bureau of Economic Research. Lichtenberg, Frank R. 2005. “Availability of New Drugs and Americans’ Ability to Work.” Journal of Occupational and Environmental Medicine 47(4): 373–80. Manski, Charles, and John Pepper. 2000. “Monotone Instrumental Variables: With an Application to the Returns to Schooling.” Econometrica 68: 997–1010. Marmot, Michael G. 2003. “Understanding Social Inequalities in Health.” Perspectives on Biology and Medicine 46(3 Suppl): S9–23. Marmot, Michael G., George Davey Smith, Stephen Stansfeld, Chandra Patel, Fiona North, Jenny Head, Ian White, Eric Brunner, and Amanda Feeney. 1991. “Health inequalities Among British Civil Servants: The Whitehall II Study.” Lancet 337(8754): 1387–93. Mathers, Colin D., and Deborah J. Schofield. 1998. “The Health Consequences of Unemployment: The Evidence.” Medical Journal of Australia 168(4): 178–82. Morris, Joan K., Derek G. Cook, and A. Gerald Shaper. 1992. “Non-Employment and Changes in Smoking, Drinking, and Body Weight.” British Medical Journal 304(6826): 536–41. ———. 1994. “Loss of Employment and Mortality.” British Medical Journal 308(6937): 1135–39. Murphy, Gregory C., and James A. Athanasou. 1999. “The Effect of Unemployment on Mental Health.” Journal of Occupational and Organizational Psychology 72(1): 83–99. Muustonen, H., P. Paakkanen, and J. Simpura. 1994. “Drinking Habits Among the Employed and Unemployed.” Nordisk Alkoholtidskrift 11(English Supplement): 21–34. National Health and Nutrition Examination Survey. Various years. Hyattsville, Md.: National Center for Health Statistics. Available at: http://www.cdc.gov/nchs/nhanes.htm (accessed March 23, 2009). Neumayer, Eric. 2004. “Recessions Lower (Some) Mortality Rates: Evidence from Germany.” Social Science of Medicine 58(6): 1037–47.

112

/

On Work and Health Ruhm, Christopher J. 1995. “Economic Conditions and Alcohol Problems.” Journal of Health Economics 14(5): 583–603. ———. 2000. “Are Recessions Good for Your Health?” Quarterly Journal of Economics 115(2): 617–50. ———. 2003. “Good Times Make You Sick.” Journal of Health Economics 22(4): 637–58. ———. 2005. “Healthy Living in Hard Times.” Journal of Health Economics 24(2): 341–63. Ruhm, Christopher J., and William E. Black. 2002. “Does Drinking Really Decrease in Bad times?” Journal of Health Economics 21(4): 659–78. Shaikh, Azeem M., and Edward J. Vytlacil. 2005. “Threshold Crossing Models and Bounds on Treatment Effects: A Nonparametric Analysis.” NBER Technical Working Paper No. 307. Cambridge, Mass.: National Bureau of Economic Research. Snyder, Stephen E., and William N. Evans. 2002. “The Impact of Income on Mortality: Evidence from the Social Security Notch.” NBER Working Paper No. 9197. Cambridge, Mass.: National Bureau of Economic Research. Snyder, Stephen E., and William N. Evans. 2003. “Lower Social Security Benefits Reduced Mortality.” NBER Digest, April 2003: 2. Sokejima, Shigeru, and Sadanobu Kagamimori. 1998. “Working Hours as a Risk Factor for Acute Myocardial Infarction in Japan: Case-Control Study.” British Medical Journal 317(7161): 775–80. Turcotte, Martin. 2008. “Time Spent with Family During a Typical Workday: 1986 to 2005.” Statistics Canada 11-008: 1–11. Turner, J. Blake. 1995. “Economic Context and the Health Effects of Unemployment.” Journal of Health and Social Behavior 36(3): 213–29. Ulrich, Cornelia M., and John D. Potter. 2006. “Folate Supplementation: Too Much of a Good Thing?” Cancer Epidemiological Biomarkers Preview 15(2): 189–93. Vistnes, Jessica Primoff, and Vivian Hamilton. 1995. “The Time and Monetary Costs of Outpatient Care for Children.” American Economic Review 85(2): 117–21. Wagstaff, Adam. 1985. “Time Series Analysis of the Relationship Between Unemployment and Mortality: A Survey of Econometric Critiques and Replications of Brenner’s Studies.” Social Science of Medicine 21(9): 985–96.

/

113

Chapter 4 Parental Pathways to Self-Sufficiency and the Well-Being of Younger Children Greg J. Duncan, Lisa Gennetian, and Pamela Morris

T

he push for antipoverty programs that promote parents’ self-sufficiency by requiring or supporting employment has been building for over thirty years, since the early 1980s. Yet increasing the self-sufficiency of single parents raises some important questions about how such strategies affect the development of their children. Most important, how do children fare when their parents increase their employment? Transitions from welfare to work may benefit children by placing them in stimulating child-care settings, creating positive maternal role models, promoting maternal self-esteem and sense of control, introducing productive daily routines into family life, and, eventually, fostering career advancement and higher earnings on the part of both parents and children. On the other hand, efforts to promote employment may overwhelm already stressed parents, force young children into substandard child care, reduce parents’ abilities to monitor the behavior of their older children, and, for those unable to sustain steady employment, deepen family poverty. This chapter contributes to the literature on parental self-sufficiency and child well-being in two ways. First, we bring a novel interdisciplinary perspective to formulating hypotheses about the pathways by which policy-induced changes in the environments in which children are embedded, both within and outside the home, facilitate or harm children’s development. These hypotheses help to organize the contradictory assertions regarding child impacts that have surrounded the debate over welfare reform. Second, we draw on a set of policy experiments to understand the effects of reforms targeting parents’ self-sufficiency on both parents and their children. The random assignment design of these evaluations provides an unusually strong basis for identifying conditions under which policy-induced increases in employ-

/

117

Making the Work-Based Safety Net Work Better

ment among low-income and mostly single parents can help or hurt young children’s achievement. Going beyond the usual documentation of experimental policy impacts, we test hypotheses regarding the pathways by which these kinds of policies affect child well-being. We do this by reviewing evidence on the relative importance of exogenous changes in parents’ income, education, parenting, employment, and child care induced by random assignment on children’s development. Thus, we bring the rigor of experimental variation to a literature that has relied exclusively on methodologies that exploit naturally occurring variation in employment or welfare transitions (for example, see work by Chase-Landsdale et al. 2003). By focusing on the influence of these pathways on children’s development, we provide a solid evidentiary base for policy recommendations regarding the design of the next generation of work-support experiments. The data and analysis in our chapter draw on research conducted as part of the Next Generation Project, a collaborative effort involving researchers at MDRC and several universities.1 Unique to this research is its synthesis of results from a number of experimental studies launched in the late 1980s and early 1990s on how programs aimed at increasing the self-sufficiency of low-income parents can affect the development of children. Our chapter concentrates on the effects of these self-sufficiency policies on the achievement of young children, the group for whom we have the strongest data and can make the clearest policy recommendations. To preview our results, we find considerable evidence for the importance of increased family income and center-based child care as key policy pathways linking policies promoting parental self-sufficiency to child achievement. Increasing employment without concurrent increases in income does not have similar positive influences on children’s development. Additionally, we find very little evidence that these policies affect child well-being through either maternal mental health or parenting. We conclude with a discussion of the costs and benefits of some self-sufficiency policies and recommendations about how children’s development can be promoted via policies aimed at the self-sufficiency of their parents.

BACKGROUND Much developmental research is consistent with the hypothesis that policies targeting low-income parents’ employment and income may affect children’s development. We outline a conceptual framework involving policy impacts on child well-being to interpret the pattern of impacts found in our experimental studies. Because the primary targets of the welfare and employment programs in this review are parents’ employment and income, we focus our brief review on nonexperimental studies of links between changes in parents’ economic behavior and children’s developmental outcomes. We summarize separately the literatures on employment and income for the sake of convenience but recognize that the two are conflated as employment leads to increased earnings and can contribute to family economic resources and consumption. 118

/

Parental Pathways to Self-Sufficiency

Employment In families headed by low-income single mothers, simple correlations between maternal employment and children’s cognitive and social development are positive, but most of these associations can be explained by differences between mothers who are employed and those who are not in such preexisting characteristics as education, number of children, and prior welfare experience (Chase-Lansdale et al. 2003; Harvey 1999; Huston 2002; Vandell and Ramanan 1992; Zaslow and Emig 1997). The effects of maternal employment on children’s development also appear to depend on the quality, extent, and timing of employment; longer hours, lower wages, and unpredictable schedules are often associated with unfavorable child outcomes.

Income Evidence on the effects of income consistently shows negative associations between poverty and children’s cognitive development. Debates continue about the causal role of income per se, as opposed to other correlates of poverty, and whether increased income can reverse the detrimental effects of poverty (Blau 1999; Bradley and Corwyn 2002; Duncan and Brooks-Gunn 1997; Duncan et al. 1998; Mayer 1997; McLoyd 1998; McLoyd, Aikens, and Burton 2006; Morris and Gennetian 2002). By emphasizing both the role and the distribution of material and social resources, economic and psychological theory suggests various mechanisms by which parents’ employment and income (Becker 1981; Bergstrom 1997; Coleman 1988) and family psychological processes (Chase-Lansdale and Pittman 2002; McLoyd 1990, 1997, 1998; McLoyd et al. 1994) affect children. Economic theories propose that employment and income affect children’s social behavior and academic achievement by influencing the goods parents purchase for children (books, toys, and child care) and the time parents spend with them. However, recent analyses of expenditure patterns among single mothers affected by welfare reform policies show increased allocation for items that facilitate work such as transportation, food eaten outside of the home, and adult clothing and little change in expenditure on children’s enrichment activities, learning, or home environment (Kaushal, Gau, and Waldfogel 2006). In contrast, psychological theory emphasizes the ways in which employment and income may affect parental emotional well-being (stress and depression). These changes affect parenting behavior and, in turn, children’s development.

Differences Across the Childhood Age Span Although we concentrate on children in the preschool and early elementary school years, it is important to realize that policy impacts can and do vary considerably across child age groups. For example, some studies point to the particular /

119

Making the Work-Based Safety Net Work Better

vulnerability of infants whose mothers are in the labor force full-time (BrooksGunn, Han, and Waldfogel 2002; Waldfogel 2006). The limited evidence for adolescents in low-income families is mixed. Some studies suggest no marked effect of maternal employment on delinquency and substance abuse (Gottfried and Gottfried 1994; Hillman and Sawilowsky 1991; Paulson 1994; Vander Ven et al. 2001), whereas others show positive associations with aspects of socio-emotional development (Chase-Lansdale et al. 2003; Richards and Duckett 1994; Muller 1995) and negative consequences for school achievement (Bogenschneider and Steinberg 1994). The effects of income may also vary across stages of childhood, but the patterns are not entirely clear. Some investigations show that associations between income and later educational attainment are strongest for preschool children (Duncan and Brooks-Gunn 1997; Duncan et al. 1998). Other investigations show that poverty in the elementary school years predicts achievement at least as well as earlier poverty (National Institute of Child Health and Human Development, Early Child Care Research Network 2006). Increases in family income produced by the Income Maintenance experiments had some positive effects on the school performance of elementary school children but not on the performance of adolescents (Salkind and Haskins 1982). All agree, however, that children living in chronic poverty are more likely to have low achievement than are those in transitory poverty at any age (National Institute of Child Health and Human Development, Early Child Care Research Network 2006). The full list of pathways by which an employment-promoting policy change might affect children’s achievement includes not only changes in family income but also changes in child care, maternal education and mental health, parenting, and other aspects of the home environment. Past literature investigating these pathways has been based almost exclusively on nonexperimental data. Studies based on a number of welfare reform experiments conducted in the 1990s have investigated several of these pathways but never more than one pathway per paper. We look across studies to synthesize lessons from the experiments about which of these pathways appear most important in explaining how employment policies affect child well-being.

EXPERIMENTS USED FOR THIS ANALYSIS The analyses conducted under the Next Generation Project are based on seven random-assignment studies that together evaluate the effects of thirteen employment-based welfare and antipoverty programs in the United States and two Canadian provinces. All told, these studies provide information on 10,664 children between our focal ages of two and five, when their intervention programs first began, who are primarily from single-parent families. All of the studies began in the early to mid-1990s (prior to 1996) and were designed to estimate the effects on low-income families and children of programs aimed at increasing parental employment. Many of these evaluations were implemented under

120

/

Parental Pathways to Self-Sufficiency

waivers of the rules governing Aid to Families with Dependent Children (AFDC), the welfare system that was replaced in 1996 by Temporary Assistance for Needy Families (TANF). Although most of the studies were under way by 1996, they were capable of testing the effects of many program features that have since been implemented by the states under TANF. (Table 4A.1 provides further detail on the studies included in this chapter.) The great contribution of these studies derives from their design, in which participants were randomly assigned to a program group that received the experimental policy package or to a control group that continued to live under the existing policies. In all but one study, parents were applying for welfare or renewing eligibility when they were randomly assigned. The exception is the New Hope program, for which all low-income adults living in two Milwaukee neighborhoods could volunteer; both program- and control-group parents remained eligible for public benefits and were subject to Wisconsin’s evolving welfare rules. Although various packages of policies were tested, we highlight the policy distinction between earnings supplement policies, which are designed to make work pay by providing cash supplements outside the welfare system or allowing parents to keep part of their welfare grant as their earnings increase, and mandatory employment services and time-limited programs, which attempt to boost work through the use of services, sanctions, and time limits. The service component of these programs mandates participation in education, training, and job-search assistance activities. Random assignment provides a strong foundation for assessing causal impacts of the welfare and employment treatment packages. At the same time, the treatments and locations in these experiments represent neither the full range of TANF programs implemented by states nor the variety of macroeconomic conditions, both good and bad, that states currently face or are likely to face in the next decade. All of the Next Generation Study programs were developed prior to the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) legislation. Some investigations were designed in response to the 1988 Family Support Act; others were tests that states chose to perform in anticipation of changes in the federal welfare law. Several mandated that recipients use employment services offered, but only two had time limits. A number of studies included earnings disregards, although in several cases the generosity of the supplements exceeded that of post-1996 state policies. Although they do not represent all of the provisions of the 1996 law, the diversity of the programs provides an opportunity to test whether child outcomes differed by the variations in policies evaluated. Briefly, the Next Generation studies are strong in internal validity but weaker in external validity for the current policy context, a point to which we return in the concluding section. Moreover, with the exception of New Hope, the Next Generation studies tell us nothing about the effects of policies on those who do not apply for or rely on the welfare system (because of the more stringent nature of the provisions, for example). To estimate average effects across studies we pooled microdata from all of the studies and estimated differences between the program and control groups on

/

121

Making the Work-Based Safety Net Work Better

follow-up measures, using OLS regression analyses.2 We report the impact of the programs as the difference between the program- and control-group levels, controlling for baseline demographic characteristics of families and children, as well as the time between random assignment and the outcome measurement. Our sample includes up to 18,667 child observations taken from 10,664 children living in 9,258 families in the seven studies.3 Children ranged in age from two to five at the point of random assignment. Children’s achievement was assessed at least two and sometimes as long as seven years after parents entered the programs, meaning that most children were in school when we assessed their achievement (across the studies, the youngest child was four years old at the time of the followup and the oldest child was twelve). The measures of achievement for young children varied and include parent reports, standardized tests, and teacher reports, with some studies having more than one type of measure. We classify children by their age at the point of their parents’ random assignment rather than by the age at the follow-up assessment. Age at random assignment better conforms to the hypothesis of developmental differences because this is the point at which children are most likely to experience the immediate effects of the intervention on their parent’s economic behavior. In the Next Generation studies, intervention impacts on parental employment and earnings typically emerged quickly after the point of random assignment, which triggered an early boost in employment. Over time, differences in employment rates diminished as parents in the control group naturally increased their employment and “caught up” to the employment rates of parents in the experimental group. Thus, outcomes measured several years later are more likely to reflect the cumulative influence of an early boost produced by the intervention on the outcome of interest (as opposed to resulting from ongoing or persistent differences in employment rates throughout the follow-up period between mothers in the experimental group and those in the control group). Random assignment marks the time at which children begin to experience the policy-induced changes in economic behavior on the part of their parents.4

PATTERNS OF ACHIEVEMENT IMPACTS We now examine the key impacts of these programs on children’s achievement, initially looking at impacts across childhood and then concentrating specifically on impacts for younger children.

Impacts Across Childhood Although we concentrate on children age two to five at the point of random assignment, we begin by summarizing the main experimental impacts on the achievement of children age two to fifteen from Pamela Morris et al. (2005).5

122

/

Parental Pathways to Self-Sufficiency

These analyses find that the effects of policies differed across the childhood age span, with positive effects on the achievement of preschool-age children and negative effects of the same policies for children entering adolescence. Additional analyses have indicated that these differences in effects across child age groups cannot be attributed to variation in family characteristics that coincide with having children in differing age groups (that is, parents of older children may have longer histories of welfare receipt or otherwise face greater risk factors than parents of younger children, or older children may be more likely to be first-born children). More specifically, for young children the analyses pointed to one particularly sensitive transition period, from the preschool to elementary school years (children ages four to five years old at baseline who enter elementary school over the follow-up period). The program effect for those children amounts to an increase of 7 percent of a standard deviation in child achievement, as measured two to five years after parents entered the programs. This is equivalent in magnitude to slightly more than a single point on a typical standardized achievement test with a standard deviation of fifteen or seven points on an SAT-type scale. By comparison, a recent random assignment experimental evaluation of Head Start (a direct intervention focused at the same age group of children) found positive shortterm effects of participating in Head Start on elementary prereading and prewriting for three- and four-year-olds equal to about 0.3 and 0.2 of a standard deviation (Ludwig and Phillips 2007). Morris, Duncan, and Clark-Kauffman (2005) report fewer effects on achievement for younger preschool children (age two to three when their parents were randomly assigned) and for children aged four to nine. At the same time, for children ages ten to eleven years old at random assignment, there were negative effects (see also Gennetian, Duncan et al. 2004 for impacts on young adolescents). These ten- and eleven-year-olds were transitioning from elementary school to middle school or junior high and into puberty when their parents entered the programs. A more complete discussion of the ways in which these programs affected older children is presented in Lisa Gennetian, Greg Duncan et al. (2004). Experimental policies that target families appear to have differing effects on children in the same family, depending on their developmental stage. The generally beneficial impacts on young children are consistent with theoretical predictions about the developmental malleability of preschool children (Shonkoff and Phillips 2000) and about the susceptibility of the early childhood period to family, as compared with peer and neighborhood, influences (Bronfenbrenner and Morris 1998, 2006; McCall 1981). Developmental theory also suggests that children in transition periods are particularly sensitive to environmental influences or changes (Graber and Brooks-Gunn 1996). Both the four- to five-year-olds and tento eleven-year-olds are in developmental transition periods, but the effects of the welfare and employment programs go in opposite directions, suggesting that the experimental policies may lead to changes in the daily environments and experiences of young children that support their transitions and to changes in experience for early adolescents that fail to support the transitions they face.

/

123

Making the Work-Based Safety Net Work Better

Policy Variation in Impacts for Younger Children Figure 4.1 shows program impacts—the standardized differences between treatment and control children—on the school achievement of children age two to five and six to nine in each of the programs in our data. Stars on top of the bars indicate program effects that are statistically significant. Figure 4.1 shows that, although none of the programs produced statistically significant improvements in children’s achievement in both age periods, impacts tend to be more consistently larger among children in programs that provided earnings supplements. As shown in table 4.1, the pooled earnings supplement program impacts for two- to five-year-olds amounted to a statistically significant 0.08 of a standard deviation, about one point on an IQ test-type scale. By comparison, the pooled impact for programs that provided only mandatory employment services or time limits without generous supplements was a statistically insignificant 0.04 (although as shown in the final column the difference in these two impacts is not statistically significant). The absence of data on six- to nine-year-old children from the National Evaluation of Welfare-to-Work study (NEWWS; a set of programs testing mandatory welfare-to-work either through labor force attachment approaches or through human capital development approaches) hampers our ability to compare across age periods. For behavioral assessments of children’s functioning, few effects are observed for either type of program for the two- to five-year-olds. Gennetian and Cynthia Miller (2002) and Aletha Huston et al. (2001) report some benefits to children’s behavior for the Minnesota Family Investment Program (MFIP) and the New Hope programs, but this does not generalize to pooled estimates from all of the earnings-supplement programs in table 4.1. None of the mandatory employment programs produced positive behavioral impacts (that is, reductions in problem behavior or increases in positive behavior). With regard to child heath, the two types of programs produce very different effects, positive for earnings-supplement programs and negative for the non-earnings-supplement programs (see Gennetian, Hill, Lopoo, and London 2008, for further discussion). With the possible exception of child health, the absence of negative child impacts from either type of program conflicts with predictions of harm to children feared by reform critics. They feared that the development of young children might be compromised by the stresses and disruptions wrought by welfare-to-work transitions. This does not appear to be the case, even in the two studies that placed time limits on welfare receipt.

PATHWAYS TO BENEFICIAL IMPACTS Because parents rather than children are the direct targets of the programs, links between experimental policies and children’s achievement are indirect, operating through changes in parents’ employment, family resources, the home or childcare environment, parent-child interactions, or parents’ anxiety, stress, and mental

124

/

T

C

0.03

0.03

0.15*

–0.01

0.18

0.04 0.01

–0.04

–0.02

Earnings-Supplements Programs

S

BC

SP

0.01

0.11**

S

B

N

SP

0.01

0.12**

Age Two to Five Age Six to Nine

L

-P

P SS

0.01

0.06

P

FT

w

FA vL sR

0.01

N

w

sA

tH

N

D C

0.10*

w

rH sG

–0.05

N

D C

Non-Earnings-Supplements Programs

N

FA rL

0.03

sG w N

FA tL

0.07

sA w N

–0.04

0.11*

Summary of Program Impacts on Children’s School Achievement

e ll P cO op FI In Fu P P H M I I l F F ew ra M M N Ru rb rb U U

0.02

/

w

sR

vH

–0.04

C

D

LA

0.00

0.07

Source: Authors’ compilation. Notes: SSP-NB = SSP New Brunswick site; SSP-BC = SSP British Columbia site; SSP-PL = SSP Plus; NwsGrHCD = NEWWS Grand Rapids, Mich., site, human capital development group; NwsGrLFA = NEWWS Grand Rapids, Mich. site, labor-force attachment group; NwsRvLFA = NEWWS Riverside, Calif. site, labor-force attachment group; NwsRvHCD = NEWWS Riverside, Calif., site, human capital development group; NwsAtLFA = NEWWS Atlanta, Ga. site, labor force attachment group; NwsAtHCD = EWWS Atlanta, Ga. site, human capital development group UrbMFIPFull = MFIP urban counties, full program group; UrbMFIPIncO = MFIP urban counties, income incentives only group; Rura1MFIP = MFIP rural counties; CT = Connecticut’s Jobs First sites; FTP = Florida’s Family Transition Program sites; LA = Los Angeles Jobs–First Greater Avenues for Independence sites. * p < 0.1; ** p < 0.05; *** p < 0.01 (two-tailed).

−0.10

−0.05

0.00

0.05

0.10

0.15

0.20

FIGURE 4.1

Effect Size of Impact on Achievement

Making the Work-Based Safety Net Work Better TABLE 4.1

/ Impacts on Developmental Outcomes of Children Age Two to Five at Time of Study Entry

Variables Achievement

Total problem behavior Parent report

Teacher report

Externalizing behavior Parent report Teacher report

P-Value of Differences Between Program Models

All Programs

EarningsSupplement Programs

Non-EarningsSupplement Programs

0.05*** (0.02) [7; 18,667]

0.08*** (0.02) [4; 8,9411

0.04 (0.03) [3;9,726]

ns

–0.01 (0.02) [6; 11,256] –0.04 (0.04) [3; 2,034]

–0.01 (0.03) [4; 5,843] –0.03 (0.08) [2; 653]

–0.02 (0.03) [2; 5,413] –0.04 (0.06) [1; 1,381]

ns

–0.02 (0.02) [6; 11,273] –0.06 (0.04) [3; 2,079]

0.00 (0.03) [4; 5,851] –0.06 (0.08) [2; 652]

–0.04 (0.03) [2; 5,422] –0.06 (0.05) [1; 1,427]

ns

ns

ns

(Table continues on p. 127.)

health. All of these indirect pathways have been supported in the nonexperimental literature (McLoyd, Aikens, and Burton 2006). More specifically, a large body of research suggests that poverty not only limits the resources that parents can provide but also increases parental stress and negative parenting practices (Bradley and Caldwell 1984; McLoyd et al. 1994; Smith, Brooks-Gunn, and Klebanov 1997; Sugland et al. 1995). In studies of parental job loss, parents who reacted with punitive, inconsistent parenting had children who experienced psychological distress and displayed problem behavior (Elder 1974, 1979; Elder, Liker, and Cross 1984; McLoyd 1990, 1998; McLoyd et al. 1994). A conceptual model of these possible mediating pathways is shown in figure 4.2. The first question in testing the mediating pathways is whether our experimental programs had any effect on any of our hypothesized mediators.

Income and Employment Our first consideration is the direct targets of these welfare and employment programs—parents’ employment and income. Finding that the magnitude of program impacts on, say, parents’ employment is roughly proportional to impacts on

126

/

Parental Pathways to Self-Sufficiency TABLE 4.1

/ (Continued)

Variables Internalizing behavior Parent report Teacher report

Total positive behavior Parent report

Teacher report

Parent report of child’s health status

All Programs

EarningsSupplement Programs

Non-EarningsSupplement Programs

0.00 (0.02) [6; 11,259] –0.01 (0.05) [3; 1,873]

0.01 (0.03) [4; 5,821] 0.05 (0.08) [2; 649]

–0.01 (0.03) [2; 5,438] –0.05 (0.06) [1; 1,224]

0.00 (0.02) [6; 11,251] 0.06 (0.04) [3; 2,064]

0.01 (0.03) [4; 5,812] 0.11 (0.08) [2; 653]

–0.01 (0.03) [2; 5439] 0.03 (0.06) [1; 1,411]

–0.01 (0.02) [6; 11,294]

0.05** (0.02) [4; 5,594]

–0.06*** (0.02) [2; 5,700]

P-Value of Differences Between Program Models ns

ns

ns

ns

0.002***

Source: Authors’ compilation. Notes: Standard errors are in parentheses. Number of studies and total sample size are in brackets. All samples consist of children age two to five at the point of random assignment. Economic variables are measured over study follow-up; child achievement is measured at time of followup. Separate regression equations are conducted for earnings-supplement programs and non-earningssupplement programs. The regressions also include the following covariates measured at baseline: earnings in the prior year, earnings in the prior year squared, amount of time mother was on welfare, employed in prior year, mother had high school degree or equivalent, mother’s marital status, number of children in the family, age of youngest child, mother’s race or ethnicity, and whether parents’ age was less than eighteen at the time of child’s birth. Also included were the following additional covariates: study site flags (for example, NEWWS—Atlanta, NEWWS—Riverside, LA-GAIN, and so on), elapsed time between study entry and follow-up, and type of achievement report (for example, parent or test or teacher, when applicable). * p < 0.1; ** p < 0.05; *** p < 0.01 (two-tailed)

children’s achievement would support a policy pathway involving increased employment. If developmental benefits are concentrated among programs with the largest impacts on family income, then the income pathway would be supported. To assess income effects, we first compared adult outcomes for parents who participated in programs that had earnings supplements and other program models (that is, programs with mandatory employment services and a program with time limits, both without earnings supplements). Table 4.2 shows similar program

/

127

Child Care

Mother's Education

Welfare Receipt

Family Income

Quality of Parent-Child Relationship

Parent's Stress and Mental Health

Family Routines

Learning Experiences in Home Environment

Child Care

Mother's Employment

Health

Academic Achievement

Social Behavior and Behavior Problems

Source: Authors’ compilation. Notes: This is not intended to be a comprehensive model detailing all pathways between welfare and employment policies and outcomes for children. Direct, indirect, reciprocal, and moderating relations are not presented.

Enhanced Child-Care Assistanace

Time Limits

Mandatory Employment Services • Labor Force Attachment (LFA) • Human Capital Development (HCD)

Earnings Supplements

Mediators

Policy Targets

Children's Developmental Outcomes

/ Conceptual Model: Possible Effects of Welfare and Employment Policies on Young Children’s Developmental Outcomes

Welfare and Employment Policies

FIGURE 4.2

Parental Pathways to Self-Sufficiency TABLE 4.2

/ Impacts on Policy Targets, Mothers of Children Age Two to Five at Time of Study Entry

P-Value of Differences Between Program Models

All Programs

EarningsSupplement Programs

Non-EarningsSupplement Programs

Participation in adult education

0.02** (0.01) [6; 11,105]

–0.01 (0.01) [4; 7,352]

0.09*** (0.01) [2; 3,753]

0.000***

Employment (fraction of quarters worked per year)

0.07*** (0.01) [7; 12,537]

0.08*** (0.01) [4; 6,408]

0.06*** (0.01) [3; 6,129]

0.08*

Earnings (in $1,000s of annual income)

0.57*** (0.09) [7; 12,537]

0.54*** (0.13) [4; 6,408]

0.56*** (0.12) [3; 6,129]

ns

Family income (in $1,000s of annual income)

0.77*** (0.09) [7; 12,537]

1.29*** (0.13) [4; 6,408]

0.17 (0.12) [3; 6,129]

0.000***

Welfare income (in $1,000s of annual income

–0.27*** (0.05) [7; 12,537]

–0.16** (0.07) [4; 6,408]

–0.36*** (0.06) [3; 6,129]

0.039**

Variables

Source: Authors’ compilation. Notes: Standard errors are in parentheses. Number of studies and total sample size are in brackets. All samples consist of children age two to five at the point of random assignment. Economic variables are measured over study follow-up; child achievement is measured at time of followup. Separate regression equations are conducted for earnings-supplement programs and non-earningssupplement programs. The regressions also include the following covariates measured at baseline: earnings in the prior year, earnings in the prior year squared, amount of time mother was on welfare, employed in prior year, mother had high school degree or equivalent, mother’s marital status, number of children in the family, age of youngest child, mother’s race or ethnicity, and whether parents’ age was less than eighteen at the time of child’s birth. Also included were the following additional covariates: study site flags (for example, NEWWS—Atlanta, NEWWS—Riverside, LA-GAIN, and so on), elapsed time between study entry and follow-up, and type of achievement report (for example, parent or test or teacher, when applicable). Dollar amounts are adjusted for inflation to 2001 dollars. * p < 0.1; ** p < 0.05; *** p < 0.01 (two-tailed)

impacts on annual earnings, small impact differences in employment, and larger impact differences on income. In non-earnings-supplement programs, parents’ increased earnings were almost entirely offset by declines in welfare payments, resulting in, at best, small gains in family income (which averaged a statistically insignificant $170 per year). Programs with earnings supplements increased family income by $1,300 per year, on average (Bloom and Michalopoulos 2001). With the

/

129

Making the Work-Based Safety Net Work Better

average level of income at just over $10,000 in the control group, this represents a substantial proportionate gain for families receiving earnings supplements. Concluding that benefits to children stem entirely from parents’ income is premature, however, because earnings-supplement programs simultaneously increased income, employment, and earnings and decreased receipt of public assistance. Table 4.2’s comparative analysis across program models does not reveal the relative extent to which income, employment, or welfare receipt are driving the child achievement impacts. To isolate the contribution of increased income for young children’s achievement, we employed an instrumental-variables (IV) strategy that took advantage of both the random assignment nature of the data and the policy variation across the studies (Morris, Duncan, and Rodrigues et al. 2006). Here we use treatment status, whether randomly assigned to the experimental group or to the control group, as an instrument or predictor of income that, because of the randomassignment design, is not correlated or confounded by other characteristics of the family, children, or local environment that may also influence income. In this way we are able to identify a causal effect of income that improves upon the associations commonly estimated in the literature to date. Key to the success of this approach is the fact that random assignment is implemented well so that assignment of parents to program and control groups serves as a source of variation in our predictor of interest (income) and that experimental treatment is unrelated to characteristics of families and children before they entered the programs (see Gennetian et al. 2005 and Gennetian, Magnuson, and Morris 2008, for more detail on this methodology with experimental data).6 A graphical representation of our instrumental-variables approach is shown in figure 4.3 (see Ludwig and Kling 2008). Each point represents deviations in mean income (in thousands of dollars) and achievement (in standard deviation units) for either the treatment or control groups in each of the programs.7 If income matters for child achievement, we would expect that the treatment group or site combinations with the biggest positive income deviations should also have the biggest positive achievement deviations. When a trend line is fit through these twenty-eight points, the slope of the line (0.06) is equal to the IV estimate of the effect of income on child achievement. To interpret this effect: Each $1,000 increase in annual income sustained on average across two to five years of follow-up boosts child achievement by 6 percent of a standard deviation.8 The inclusion of employment and welfare receipt as additional endogenous variables in this model does not change the size of the income effect. Neither employment nor welfare receipt comes close to being statistically significant in the multiple-endogenous-variable IV models (Morris, Duncan, and Rodrigues 2006).

Education Table 4.2 also shows increases in participation in education in the non-earningssupplement programs. This comes from the human capital–development (HCD) 130

/

FIGURE 4.3

/

Individual Study Achievement Means by Income Means Slope = 0.0559

0.08 Mean Child Achievement (Normalized)

A

B

0.06 C 0.04 D I 0.02

H M –1 Q

−1.5

–0.5

0 R

E

N 0

–0.02

S V

G F

O T

J K

L Z

P 0.5

1

1.5

U

X

W –0.04 Y –0.06 Z AA

–0.08

BB

–0.1 Mean Annual Income (Normalized) Source: Authors’ compilation. Notes: A = Self-Sufficiency Project (SSP)—British Columbia (BC), experimental group B = SSP—New Brunswick (NB), experimental group C = Florida’s Family Transition Program (FTP), experimental group D = National Evaluation of Welfare-to-Work Strategies (NEWWS)—Riverside, control group E = NEWWS—Atlanta Human Capital Development (HCD), experimental group F = NEWWS—Riverside Labor Force Attachment (LFA), experimental group G = Minnesota Family Investment Program (MFIP)—Urban Full Program, experimental group H = Los Angles Jobs-First Greater Avenues for Independence (LA-GAIN), experimental group I = NEWWS—Grand Rapids LFA, experimental group J = SSP—Plus, experimental group K = New Hope (NH), experimental group L = Connecticut’s Jobs First (CT), experimental group M = MFIP—Rural , control group N = NEWWS—Grand Rapids, control group O = NEWWS—Atlanta LFA, experimental group P = MFIP—Rural, experimental group Q = MFIP—Urban, control group R = CT, control group S = NH, control group T = NEWWS—Grand Rapids HCD, experimental group U = LA-GAIN, control group V = SSP—Plus, control group W = NEWWS—Atlanta, control group X = MFIP—Urban Incentives Only (IO), experimental group Y = FTP, control group Z = SSP—BC, control group AA = SSP—NB, control group BB = NEWWS—Riverside HCD, experimental group /

131

Making the Work-Based Safety Net Work Better

approach in the National Evaluation of Welfare-to-Work Strategies (NEWWS) sites, which focused welfare recipients first on education and training prior to employment. Other NEWWS mothers were randomized into a labor-force attachment (LFA) approach that encouraged recipients to take jobs quickly. Although one might expect that programs targeting maternal human capital would be beneficial for children, HCD program impacts on child achievement were not statistically significant. Katherine Magnuson (2003) found that mothers in the HCD program stream acquired very little additional education—just over two months on average—so the magnitude of change in human capital was in fact very small. To further explore whether education had the potential to boost children’s achievement, Magnuson (2003) estimated IV models by relating impacts on completed maternal schooling to impacts on child achievement. She found substantial effects, with every ten-month increase in maternal schooling associated with an increase in child achievement of about a quarter of a standard deviation. Thus, human capital–development programs for mothers could benefit children, but only if mothers spend time to acquire the new skills.

Child Care The mediating role of child care, in particular center-based care arrangements, was investigated by Gennetian et al. (2007). Center-based care is defined as any licensed or regulated care that takes place in a group setting; it includes child-care centers, Head Start, or other early education settings, as well as organized beforeand after-school programs. Home-based care means care by relatives or others in the child’s own home or another person’s home. Although all of the programs increased parents’ employment and the use of paid child care (Gennetian, Crosby et al. 2004), the type of child care affected depended on the program model and its provision of expanded child-care assistance (see table 4.3). Perhaps owing to some combination of policy impacts on income and employment, programs with earnings supplements led to increased use of center-based over home-based child care. The reverse is true for programs without such supplements. In addition, program effects on child care were differentiated by the extent to which programs included expanded child-care assistance, a feature of a few earnings-supplement programs and one non-earnings-supplement program (results not shown; see Crosby, Gennetian, and Huston 2005). Programs with such expanded assistance increased the use of center-based programs more than the use of home-based care arrangements. By contrast, programs that did not offer such expanded assistance increased home-based care rather than center-based care. Past research has shown modest positive achievement effects of center-based care (for example, National Institute of Child Health and Human Development, Early Child Care Research Network and Duncan 2003) for middle-income children, and intensive center-based programs such as Perry Preschool and Abecedarian provide enduring benefits for low-income children. Gennetian, Danielle Crosby et al. (2004) lacked information about the quality of the care arrangements; however, center-based 132

/

TABLE 4.3

/ Impacts on Family Mediators, Children, and Families of Children Age Two to Five at Time of Study Entry

P-Value of Differences Between Program Models

All Programs

EarningsSupplement Programs

Non-EarningsSupplement Programs

Only center-based care (percentage)

1.02 (0.84) [6; 8,399]

2.91*** (0.98) [4; 4,866]

–1.90 (1.47) [2; 3,533]

0.011**

Only home-based care (percentage)

2.64*** (0.97) [6; 8,485]

0.76 (1.36) [4; 4,903

5.33*** (1.35) [2; 3,582]

0.024**

HOME cognitive stimulation

0.00 (0.03) [5; 5,834]

0.00 (0.04) [3; 2,276]

0.00 (0.03) [2; 3,558]

ns

Maternal parenting warmth

0.00 (0.02) [6; 7,566]

–0.01 (0.03) [4; 4,156]

0.03 (0.03) [2; 3,410]

ns

Maternal parenting harshness

0.01 (0.02) [4; 2723]

0.00 (0.03) [3; 2,094]

0.07 (0.05) [1; 629]

ns

Maternal parenting aggravation

0.00 (0.02) [5; 8,151 ]

0.00 (0.04) [3; 2,471 ]

–0.01 (0.03) [2; 5,680]

ns

Mother at risk for depressive symptoms (percentage)

0.87 (0.95) [6; 11,445]

2.26* (1.36) [4; 5,893]

–0.45 (1.31) [2; 5,552]

ns

Married (percentage)

–0.39 (0.60) [7; 12,497]

–0.22 (0.85) [4; 6,399]

–0.60 (0.84) [3; 6,098]

ns

Married or cohabiting (percentage)

0.01 (0.76) [7 12,485]

0.04 (1.09) [4; 6,398]

–0.13 (1.07) [3; 6,087]

Variables

Source: Authors’ compilation. Notes: Standard errors are in parentheses. Number of studies and total sample size are in brackets. All samples consist of children age two to five at the point of random assignment. Economic variables are measured over study follow-up; child achievement is measured at time of follow-up. Separate regression equations are conducted for earnings-supplement programs and non-earningssupplement programs. The regressions also include the following covariates measured at baseline: earnings in the prior year, earnings in the prior year squared, amount of time mother was on welfare, employed in prior year, mother had high school degree or equivalent, mother’s marital status, number of children in the family, age of youngest child, mother’s race or ethnicity, and whether parents’ age was less than eighteen at the time of child’s birth. Also included were the following additional covariates: study site flags (for example, NEWWS—Atlanta, NEWWS—Riverside, LA-GAIN, and so on), elapsed time between study entry and follow-up, and type of achievement report (for example, parent or test or teacher, when applicable). * p < 0.1; ** p < 0.05; *** p < 0.01 (two-tailed)

/

133

Making the Work-Based Safety Net Work Better

settings may be beneficial for low-income children because they are, on average, of higher quality than the home-based arrangements that low-income parents use (Coley, Chase-Lansdale, and Li-Grining 2001; Dowsett et al. 2008; National Institute of Child Health and Human Development, Early Child Care Research Network 1997).9 By and large, programs that increased the use of center-based child care tended to have positive impacts on children’s achievement. IV estimates showed that the use of center-based care, as opposed to care in someone’s home, during a child’s preschool years has a positive effect on school achievement in the early grades of elementary school (Gennetian et al. 2007). Effect sizes are modest but comparable to those for income—an increase in 0.10 of probability of being exclusively in center-based care during the preschool years increases achievement by about 10 percent of a standard deviation. These analyses also show a positive effect of increased income, but once centerbased care is included in the model, the positive income effect on children’s achievement is substantially reduced, suggesting that some of the income effect may be a result of using center-based care. Because Gennetian et al. (2007) could not separate program impacts on income and on center-based care (indeed, they find that income predicts center-based care, controlling for employment), it seems reasonable to conclude that both may contribute to improvements in children’s achievement. Comparable analyses were also conducted examining the effects of center-based care on parent and teacher reports of externalizing behavior problems (Crosby et al., forthcoming). These results show that, although center-based care is positively associated with externalizing behavior in OLS estimates, IV estimates indicate lower externalizing levels, as rated by parents and teachers, for children who were in center-based care arrangements exclusively compared to those children who were not in care. The positive association in the OLS estimates may reflect a battery of factors related to selection of children in certain types of care environments whereas the IV estimates isolate the effect of increased centerbased care in response to these policies independent of personal preferences, cost considerations, or other factors that affect child-care decisions.

Parenting, the Home Environment, and Maternal Mental Health Changes in the home and parenting environment might also be pathways from the policies to outcomes for children. On the basis of earlier theory and literature (for example, McLoyd et al. 2006), we expected that increased income might improve the quality of learning experiences provided in the home, reduce parents’ stress and depression, and improve the quality of parenting behavior. Surprisingly, across all of the Next Generation studies, there were few effects of any type of programs on available measures of parenting, depression, and the home environment (table 4.3). Nor did we find impacts on marriage and cohabitation. The absence of effects on the home environment and parenting may be due to the fact that the increases in income are accompanied by increases in employ-

134

/

Parental Pathways to Self-Sufficiency

ment, offsetting any potential benefits of reduced financial strain with increased time pressure. Moreover, if much of the increased income is going to purchase child care and, at least as evidenced by one study that collected such information, to clothing and food for children, it may simply not be enough to reduce the strain of low income and result in measurable changes in parents’ behavior and emotional well-being. Whatever the reason, policymakers hoping for dramatic improvements in children’s home environments and parenting from work itself can find no support from these policy experiments.

ARE EARNINGS SUPPLEMENT PROGRAMS GOOD POLICY? Although most of the experimental programs reviewed here boosted employment and reduced welfare expenditures, and some promoted children’s school achievement, it remains to be established whether any of them would pass a benefit-cost policy test. Cost-benefit analyses typically distinguish costs borne by and benefits accruing to participants and the costs and benefits for the rest of society (often referred to as “taxpayers”). Summing across these two groups provides estimates of total social costs and benefits. Although policymakers sometimes choose to focus only on taxpayers’ costs and benefits, economic logic stresses the need to compare total resource costs, regardless of the degree to which they are borne by taxpayers or the participants, and total benefits, regardless of whether they are enjoyed by taxpayers or only by participants. In the case of our work-support programs, the most obvious benefit is the greater economic productivity of participants as reflected in their higher earnings. The benefits of higher earnings accrue mostly to program participants (who receive both the higher earnings and often additional income from the Earned Income Tax Credit [EITC] and child tax credit), although if earnings are high enough taxpayers will benefit from the added taxes paid. In the case of the earnings supplement programs, the benefits associated with higher achievement and, in some cases, better behavior among children in the program groups need to be factored in as well. The evaluations of three of the earnings supplement programs—the Minnesota Family Investment Program (MFIP), the Canadian Self-Sufficiency program (SSP), and Milwaukee New Hope—provided accounting of costs and benefits, although only in the case of New Hope was there an attempt to monetize the value of the child benefits. The Canadian SSP program produced some of the largest earnings gains. Its cost-benefit study (Michalopoulos et al. 2002) showed that the program’s fiveyear costs to taxpayers (−$2,691) were more than offset by its earnings-driven gains to participants (+$5,256), producing a net gain to society of +$2,565. Labor-market impacts were not as large in the case of MFIP, and, in contrast to SSP, the program increased rather than reduced welfare use (Miller et al. 2000). In the case of long-term welfare-recipient single parents, the $3,208 earnings gains over a five-year period more than offset the program’s operating costs, although

/

135

Making the Work-Based Safety Net Work Better

these gains were dwarfed by the nearly $7,000 in welfare and EITC payments transferred from taxpayers to participants. The result was a net gain from a society perspective but a net loss to taxpayers. New Hope produced modest impacts on earnings, in part because about onethird of its participants already had full-time jobs when they signed up for the program. The annual-earnings advantage of New Hope participants relative to controls amounted to $497 (Duncan, Huston, and Weisner 2007). Although the earnings impacts were twice as high for the participants who were not working at the start of New Hope, and more than $2,500 per year for the subset of participants with some but not too many employment barriers, the overall productivity estimate of $497 is most appropriate for a program-wide estimate of benefits. New Hope’s collection of work supports, if implemented today, would be costly⎯$284 per year, on average, for its earnings supplements, $875 for the health insurance subsidies, and $470 for the child-care subsidies (Bos et al. 2007). The administrative costs of running a New Hope program would also be substantial, amounting to a $1,717 per year per participant. All told, the labor market–related benefits of New Hope do not come close to covering its $3,308 costs. What about the child benefits? The benefit-cost calculation might be positive if, for example, the benefits were expanded to consider the long-term effects of center-based care on children’s development. Johannes Bos et al. (2007) rely on a method for converting teacher-reported achievement gains into the monetary value of the future earnings gains the children would likely receive. They estimate the annual discounted value of the future earnings gains to be $1,295, of which $1,036 is enjoyed by the children themselves and $259 accrues to taxpayers in the form of higher taxes paid. Teachers also reported that, relative to control-group boys, boys in the New Hope program showed much more positive behavior, exhibited fewer behavior and disciplinary problems, and were more compliant and less hyperactive in classrooms. Based on published estimates of the dollar value to taxpayers of saving a youth from serious crime and drug abuse and from dropping out of high school, Bos et al. (2007) calculate the taxpayer value of saving a high-risk New Hope child to be $2.4 million, with crime-victim and criminal-justice costs accounting for the bulk of this total.10 They estimate that about one in sixteen New Hope boys would have to transition out of the “high-risk” category to pay for the entire taxpayer cost of the New Hope program, and fewer than one in thirty-one such transitions are required to pay for the entire societal costs of the program. These calculations illustrate that the child impacts in self-sufficiency programs, even those that provide supplementary support for health insurance and childcare subsidies, can be an important part of their potential value to society.

SUMMARY AND DISCUSSION Taken together, we find that parents’ income and children’s child-care arrangements appear to be key pathways for explaining the beneficial effects of some welfare and employment policies on young children (see figure 4.4). Programs that 136

/

Parental Pathways to Self-Sufficiency

increase income and use of center-based child care are most able to improve children’s achievement, measured a few years after program onset. There is little evidence from the policy experiments that increases in employment or reductions in welfare by themselves produce detectable impacts on young children’s achievement. Programs targeting maternal human capital have the potential to improve outcomes for children, but only if mothers acquire enough of it. Among the mediators listed in figure 4.4, center-based child care stands out as the primary way in which a policy targeted to adults allows them to translate their higher income into investments that can improve the well-being of children. There are no detectable differences in material hardship, food security, or child neglect to the extent that these aspects of day-to-day living can be measured from parent reports. While many studies have attempted to estimate the effects of income on children’s development, few rely on experimental or quasi-experimental variation in income. As such, these findings contribute to ongoing debates on how (and, even whether) policy-induced increases in family income would be spent in ways that would boost the achievement of children (Blau 1999; Duncan and Brooks-Gunn 1997; Haveman and Wolfe 1995; Jacob and Ludwig 2006; Mayer 1997). The patterns of effects aligning income, center-based care, and children’s developmental outcomes are compelling and informative for guiding future policies, particularly as they add confirmatory evidence to the numerous positive associations commonly documented in the literature to date. The extent to which such effects hold up under slightly varying economic contexts, evolving and more stringent welfare-towork policies, or compositional changes among low-income workers is an open question that we think should guide future research in efforts to inform policy. Although these analyses provide a great deal of information, many questions remain. First, current law permits states to require mothers to participate in workrelated activities when their children are very young infants. The experiments in our analysis includes few infants younger than a year old, but other research investigating the effects of maternal employment on such young children raises questions about possible negative effects of full-time employment (Waldfogel 2006). A National Academy of Science panel specifically recommended that welfare policies should not require full-time maternal employment when children were less than a year old (Smolensky and Gootman 2001). Second, the policies tested had the most positive effects on preschool-age children (from about two to five years old), and it appears that these positive effects are in part due to increased income. How did higher levels of family income affect younger children, especially since some of this income was likely used to pay for work-related expenses, including child care? Data from these experimental studies have little to offer about consumption or expenditure patterns to inform our thinking. The one pathway that is supported is children’s centerbased child-care environments. At the same time, we lack information on the quality of care children received. Policies for working parents’ child-care assistance are typically separated from policies designed to use early educational settings to promote school readiness, even though the same children are af/

137

Center-Based Care

Enhanced Child-Care Assistanace

Source: Authors’ compilation.

Mother's Education

Welfare Receipt

Family Income

Center-Based Care

Mother's Employment

Time Limits

Mandatory Employment Services • Labor Force Attachment (LFA) • Human Capital Development (HCD)

Earnings Supplements

Mediators

Policy Targets

Academic Achievement

Children's Developmental Outcomes

/ Model Illustrating Evidence on Direct and Mediating Relations Between Welfare and Employment Policies and Young Children’s Academic Achievement

Welfare and Employment Policies

FIGURE 4.4

Parental Pathways to Self-Sufficiency

fected by both. Research on integration of these services would inform both types of policy. Contrary to our expectations, parents’ psychological well-being and parenting practices (for example, warmth, patterns of discipline) did not appear as a pathway for program effects on young children’s achievement. One might question the quality of the parenting measures (based as they are on self-reports rather than direct observation), but it is also likely that the programs did not have large impacts on these psychosocial aspects of parenting. If one defines parenting more broadly as “family management” (Duncan and Chase-Lansdale 2001), then parents’ choices about child care, living environments, schools, and other environments for their children would be included. Increased resources might affect these choices. In fact, nonexperimental investigations suggest that investments in children’s environments are better predictors of cognitive and academic skills than parenting warmth and control (Yeung, Linver, and Brooks-Gunn 2002). Findings from this set of policy experiments point to ways in which employment and income-support programs can affect children’s development. Policies that increase employment and income positively affect the achievement of children who were preschoolers when their mothers increased their work effort. Income and center-based care appear to be key ways in which these effects occurred. Maternal education has the potential to play a role but not unless mothers participate in and complete more schooling than what was produced by the programs we considered. Evidence in support of parent-child interactions and maternal mental health is limited, partly because of poor measurement but also partly because these programs did not alter the psychosocial aspect of parenting (but rather family management aspects that affect child-care decisions, for example). The studies examined here include policies that are comparable to the most generous policies currently in effect. The maximum value of the EITC more than doubled during the 1990s, providing an increased earnings supplement for all lowincome workers at a level similar to those in the generous policies examined here. In addition, most states have implemented an “enhanced earnings disregard” as part of their welfare reform strategy, which allows parents to earn more without penalization. In a handful of states, the enhanced earnings disregards are relatively generous. A welfare recipient in Connecticut, for instance, can now continue receiving all of her welfare benefits as long as she earns less than the federal poverty threshold. Relative to how she would have fared under the AFDC system, this disregard provides her with about $500 more per month in income. And California allows welfare recipients who work to keep the first $225 of their monthly earnings without having their welfare benefits reduced; beyond that point, each additional dollar of earnings reduces their benefits by only half a dollar (rather than reducing benefits by about a dollar for every dollar of earnings, as under AFDC). Our studies examining the effects of generous supplement programs are probably quite applicable to these contexts. At the same time, some enhanced disregards are not as generous as the supplements provided by the programs analyzed in this chapter. In some states, the disregard is very low, sometimes as low as 20 percent (in Alabama, for example). Also, in states with very low benefit levels (for

/

139

Making the Work-Based Safety Net Work Better

example, in West Virginia, where the welfare benefit is only $253 and the earnings disregard is 40 percent), even an enhanced earnings disregard translates into very little increase in family income. In these cases, results from programs that increase employment but not income are likely to be the most relevant benchmark. What about time limits and mandates? Only two of the studies examined included time limits, but forty states do have time limits that result in loss of benefits (Bloom, Farrell et al. 2002). Moreover, nearly all states (except for a few that are more similar to the programs we evaluated) now sanction families who are noncompliant with program rules by closing the case or taking away the entire welfare benefit, whereas the studies examined here typically sanctioned parents by the removal of the adult portion of the grant. In short, the differences in the studies we have examined and those in effect today are primarily in their focus on benefit reduction policies. Thus, there may be consequences of income loss and benefit termination for children that are not well documented in this body of work. Notably, a further examination of Florida’s time-limit policy did not suggest harm to children of families reaching welfare time limits and having their benefits reduced (Morris and Hendra, forthcoming), providing no evidence of negative effects of time limits on children and families, at least as implemented in Florida. A key finding from the experiments is that impacts on child achievement were consistently more positive in programs that provided financial and in-kind supports for work than in those that did not. The packages of work supports were quite diverse, ranging from generous financial supplements provided alone to more comprehensive packages of financial supplements, child-care assistance, health insurance, and even temporary community service jobs. Although more costly than the “work first” approach taken by the programs with mandatory employment services only, two of the programs with earnings supplements had costs within the range of some of the actual welfare reform packages implemented by states in response to the 1996 legislation. Relative to the AFDC program, the average yearly cost for a participant in a program with mandatory employment services ranged from savings of $255 to a cost of $1,595. The annual taxpayer costs per participant of the earnings supplement programs ranged from $2,000 to $4,000 above the costs of the AFDC program. At the same time, increased taxes, reductions in reliance on public assistance, and as yet unquantified taxpayer savings from the improvement in children’s achievement return at least a portion of these costs. These findings suggest that policymakers face a choice when deciding which welfare reforms are best for children. They can increase parental self-sufficiency, provide few benefits to children, and save government money with mandatory employment service programs. Or, at greater taxpayer cost, they can use earnings supplements to increase parental employment, raise family income, and provide benefits to children. Clearly, welfare policies can affect and improve the wellbeing of children if states or the federal government choose to spend additional money on work supports in the context of their welfare programs. Our investigation of the mediating pathways by which these welfare policies occurred suggest that, for younger children, center-based care is a worthwhile target of influence, via encouraging parents to take up such care through subsidies, increased income, or other policy levers. 140

/

√

Los Angeles Jobs-First Greater Avenues for Independence (GAIN) Self-Sufficiency Project (SSP)

√

√

√

√

√

√

√

√

√

√

√

√

√

1992 36 months 54 months

1996 24 months

1994 24 months 60 months

1991 24 months 60 months

1994 36 months

1996 36 months 1994 48 months

Freedman et al. (2000) Morris and Michalopoulos (2000)

Gennetian and Miller (2000) Hamilton et al. (2002) and McGroder et al. (2000) Bos et al. (1999)

Bloom et al. (2002) Bloom et al. (2000)

When Study Began and Length of Primary Follow-Up Source(s)

Source: Authors’ compilation. Notes: All sites used a random-assignment design that consisted of one or more program group and a control group. The control group in each case was the traditional welfare system in place at the time of the study (typically, AFDC).

New Brunswick British Columbia

Los Angeles County

Key Policy Features Tested Generous Mandatory Expanded Earnings Employment Time Child-Care Supplements Services Limits Assistance

National Evaluation of Welfare-to- Atlanta, Ga.; Grand Work Strategies (NEWWS) Rapids, Mich.; Riverside, Calif.; and Portland, Ore. New Hope Project Milwaukee, Wis.

Minnesota Family Investment Program (MFIP)

Family Transition Program (FTP)

New Haven and Manchester, Connecticut Escambia County, Florida Seven counties in Minnesota

Sites

Descriptions of the Studies

Connecticut Jobs-First Evaluation

Study

Appendix TABLE 4A.1 /

APPENDIX

Making the Work-Based Safety Net Work Better

This chapter was completed as part of the Next Generation project, which examines the effects of welfare, antipoverty, and employment policies on children and families. The work described in the chapter was funded by the Next Generation Project funders: the David and Lucile Packard Foundation, William T. Grant Foundation, the John D. and Catherine T. MacArthur Foundation, the Annie E. Casey Foundation, and by grant R01HD045691 from the National Institute of Child Health and Human Development. The results reflect the views of the authors and not those of any of the funders. We thank Robert Haveman, Carolyn Heinrich, Karl Scholz, and Barbara Wolfe for helpful comments on the conference draft and Francesca Longo for excellent research assistance. The three authors are listed in alphabetical order to reflect their equal contributions.

NOTES 1. For more information on the Next Generation Project, see http://www.mdrc.org/ NextGeneration (accessed March 1, 2009). 2. All of the studies were conducted by MDRC (Manpower Demonstration Research Corporation) and gathered data with virtually identical methods. An alternative approach to synthesis is to apply meta-analytic techniques (Lipsey and Wilson 1996) to impacts estimated from the individual studies. The overall results from pooling are identical to those obtained by meta-analysis, and pooling provides considerably more flexibility for estimating some of our models. Note that all of the pooled analyses have site-, city-, and state-specific controls as well as a variety of control variables on the background characteristics of the respondents. 3. We have 18,667 child observations on achievement. Achievement is sometimes assessed in more than one way within a wave of data and in most studies is gathered at multiple follow-up points. Most other child outcomes are measured using a single method across multiple follow-up points, so in those cases the number of child observations is usually 14,308 instead of 18,667. Exact sample sizes are noted in the tables. 4. Our regressions control for time between random assignment and the point at which the given outcome is measured. Our tests for differential program impacts for early versus later measurements lacked sufficient power to detect possible interactions. 5. Morris, Duncan, and Elizabeth Clark-Kauffman (2005) focused on children’s achievement in school in these analyses because that information was collected consistently across a large age range of children, including children between the ages of two and fifteen at the point of random assignment, who were then assessed two to five years later. 6. These models pool data across studies and estimate the following achievement model: Inc = T*S γ1 + S γ2 + X β1 + ε1 Ach = Inc λ1 + S λ2 + X β2 + ε2

4.1 4.2

Inc is family income, T is a treatment or control dummy variable, S is a vector of fifteen of the sixteen site or program dummies, X is a vector of control variables, and

142

/

Parental Pathways to Self-Sufficiency

7. 8.

9.

10.

Ach is child achievement. The interactions between treatment group assignments T and site indicators S are used as instrumental variables to isolate experimentally induced variations in income (Inc) and achievement (Ach) across sites. Additional endogenous variables, in particular employment and welfare receipt, can be added to this income-based model. Each group or site’s treatment and control means are expressed as deviations from the overall group or site mean. As might be expected, this estimated income effect is consistent with the patterns of program impacts shown in tables 4.1 and 4.2. Programs with earnings supplements increased family income for younger children by between $800 and nearly $2,200 per year, which corresponds to achievement effect sizes ranging from 5 to 12 percent of a standard deviation. That low-income families can generally access an organized care setting that is similar to the quality of care accessed by the families in these programs and that is of higher quality than home-based care settings is an underlying assumption in our analysis because we do not have information about the quality of care. Though developmental theories and research emphasize quality as the critical dimension determining whether child care will have positive or negative effects, the type of care also predicts academic skills independently of observed quality (see Fuller et al. 2002). The key reference here is Mark Cohen (1998).

REFERENCES Becker, Gary S. 1981. A Treatise on the Family. Cambridge, Mass.: Harvard University Press. Bergstrom, Theodore. 1997. “A Survey of Theories of the Family.” In Handbook of Population and Family Economics, edited by Mark R. Rosenzweig and Oded Stark. Volume 1A. Ann Arbor: University of Michigan. Blau, David M. 1999. “The Effect of Income on Child Development.” Review of Economics and Statistics 81(2): 261–76. Bloom, Dan, Mary Farrell, Barbara Fink, and Diana Adams-Ciardullo. 2002. Welfare Time Limits: State Policies, Implementation, and Effects on Families. New York: MDRC. Bloom, Dan, James J. Kemple, Pamela Morris, Susan Scrivener, Nandita Verma, and Richard Hendra. 2000. The Family Transition Program: Final Report on Florida’s Initial TimeLimited Welfare Program. New York: MDRC. Bloom, Dan, and Charles Michalopoulos. 2001. How Welfare and Work Policies Affect Employment and Income: A Synthesis of Research. New York: MDRC. Bloom, Dan, Susan Scrivener, Charles Michalopoulos, Pamela Morris, Richard Hendra, Diana Adams-Ciardullo, and Johanna Walter. 2002. Jobs First: Final Report on Connecticut’s Welfare Reform Initiative. New York: MDRC. Bogenschneider, Karen, and Lawrence Steinberg. 1994. “Maternal Employment and Adolescent Academic Achievement: A Developmental Analysis.” Sociology of Education 67: 60–77. Bos, Johannes, Greg Duncan, Lisa Gennetian, and Heather Hill. 2007. “New Hope: Fulfill-

/

143

Making the Work-Based Safety Net Work Better ing America’s Promise to ‘Make Work Pay.’ ” Hamilton Project Paper. Washington, D.C.: Brookings Institution. Bos, Johannes, Aletha Huston, Robert Granger, Greg Duncan, Thomas Brock, and Vonnie McLoyd. 1999. New Hope for People with Low Incomes: Two-Year Results of a Program to Reduce Poverty and Reform Welfare. New York: MDRC. Bradley, Robert H., and Betty M. Caldwell. 1984. “174 Children: A Study of the Relation Between the Home Environment and Early Cognitive Development in the First 5 Years.” In The Home Environment and Early Cognitive Development, edited by Allen Gottfried. Orlando, Fla.: Academic Press. Bradley, Robert H., and Robert Corwyn. 2002. “Socioeconomic Status and Child Development.” Annual Review of Psychology 53: 371–99. Bronfenbrenner, Urie, and Pamela Morris. 1998. “The Ecology of Developmental Processes.” In Theoretical Models of Human Development, edited by William Damon. Volume 1: Handbook of Child Psychology, edited by Richard M. Lerner. 5th ed. New York: Wiley. ———. 2006. “The Bioecological Model of Human Development.” In Theoretical Models of Human Development, edited by William Damon. Volume 1: Handbook of Child Psychology, edited by Richard M. Lerner. 5th ed. New York: Wiley. Brooks-Gunn, Jeanne, Wen-Jui Han, and Jane Waldfogel. 2002. “Maternal Employment and Child Cognitive Outcomes in the First Three Years of Life: The NICHD Study of Early Child Care.” Child Development 73(4): 1053–72. Chase-Lansdale, P. Lindsay, Robert Moffitt, Brenda Lohman, Andrew Cherlin, Rebekah Coley, Laura Pittman, Jennifer Roff, and Elizabeth Votruba-Drzal. 2003. “Mothers’ Transitions from Welfare to Work and the Well-Being of Preschoolers and Adolescents.” Science, no. 299: 1548–52. Chase-Lansdale, P. Lindsay, and Laura Pittman. 2002. “Welfare Reform and Parenting: Reasonable Expectations.” Future of Children 12(1): 167–85. Cohen, Mark. 1998. “The Monetary Value of Saving a High-Risk Youth.” Journal of Quantitative Criminology 14(1): 5–33. Coleman, James S. 1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology 94: S95–S120. Coley, Rebekah Levine, P. Lindsay Chase-Lansdale, and Christine Li-Grining. 2001. “Child care in the Era of Welfare Reform: Quality, Choices, and Preferences.” Policy Brief 01-4, Report of Welfare, Children, and Families: A Three-City Study. Baltimore, Md.: Johns Hopkins University. Crosby, Danielle, Chantelle Dowsett, Lisa Gennetian, and Aletha Huston. Forthcoming. “The Effects of Center-Based Care on the Problem Behavior of Low-Income Children with Working Mothers.” Developmental Psychology. Crosby, Danielle A., Lisa A. Gennetian, and Aletha C. Huston. 2005. “Child Care Assistance Policies Can Affect the Use of Center-Based Care for Children in Low-Income Families.” Journal of Applied Developmental Science 9(2): 86–106. Dowsett, Chantelle, Aletha Huston, Amy Imes, Lisa Gennetian, and Desiree Principe. 2008. “Structural and Process Features in Three Types of Child Care for Children from High and Low Income Families.” Early Childhood Research Quarterly 23: 69–93. Duncan, Greg J., and Jeanne Brooks-Gunn, eds. 1997. Consequences of Growing Up Poor. New York: Russell Sage Foundation.

144

/

Parental Pathways to Self-Sufficiency Duncan, Greg, and P. Lindsay Chase-Lansdale. 2001. “Welfare Reform and Child WellBeing.” Working Paper No. 217. Chicago: Northwestern University, Joint Center Poverty Research, Duncan, Greg, Aletha Huston, and Thomas Weisner. 2007. Higher Ground: New Hope for the Working Poor and Their Children. New York, Russell Sage Foundation. Duncan, Greg J., Wei-Jun Yeung, Jeanne Brooks-Gunn, and Judith R. Smith. 1998. “Does Poverty Affect the Life Chances of Children?” American Sociological Review 63(3): 406–23. Elder, Glen. 1974. Children of the Great Depression. Chicago: University of Chicago Press. ———. “Historical Change in Life Patterns and Personality.” In Life-Span Development and Behavior, edited by Paul B. Baltes and Orville G. Brim. Volume 2. New York: Academic Press. Elder, Glen, Jeffrey Liker, and Catherine Cross. 1984. “Parent-Child Behavior in the Great Depression: Life Course and Inter-Generational Influences.” In Life-Span Development and Behavior, edited by Paul B. Bates and Orville G. Brim Jr. Volume 6. Orlando, Fla.: Academic Press. Freedman, Stephen, Jean Tansey Knab, Lisa A. Gennetian, and David Navarro. 2000. The Los Angeles Jobs-First GAIN Evaluation: Final Report on a Work First Program in a Major Urban Center. New York: MDRC. Fuller, Bruce, Sharon Kagan, Susanna Loeb, Judith Carroll, Jan McCarthy, Gege Kreicher, Bidemi Carrol, Ginger Cook, Yueh-Wen Chang, and Susan Sprachman. 2002. New Lives for Poor Families? Mothers and Young Children Move Through Welfare Reform. Berkeley and New Haven: The Growing Up in Poverty Project. Gennetian, Lisa, Danielle Crosby, Chantelle Dowsett, Aletha Huston, and Desiree Alderson. 2007. “Maternal Employment, Early Care Settings, and the Achievement of LowIncome Children.” Next Generation Working Paper No. 30. New York: MDRC. Gennetian, Lisa A., Danielle Crosby, Aletha C. Huston, and Edward D. Lowe. 2004. “How Child Care Assistance in Welfare and Employment Programs Can Support the Employment of Low-Income Families.” Journal of Policy Analysis and Management 23(4): 723–43. Gennetian, Lisa, Greg Duncan, Virginia Knox, Wanda G. Vargas, Elizabeth Clark-Kauffman, and Andrew London. 2004. “How Welfare Policies Can Affect Adolescents: A Synthesis of Evidence from Experimental Studies.” Journal of Research on Adolescence 14(4): 399–423. Gennetian, Lisa, Heather Hill, Leonard Lopoo, and Andrew London. 2008. “Maternal Work Hours and Adolescents’ School Outcomes Among Low-Income Families in Four Urban Counties.” Demography 45(1): 31–53. Gennetian, Lisa, Katherine Magnuson, and Pamela Morris. 2008. “From Statistical Associations to Causation: What Developmentalists Can Learn from Instrumental Variables Techniques Coupled with Experimental Data.” Developmental Psychology 44(2): 381–94. Gennetian, Lisa, and Cynthia Miller. 2000. Reforming Welfare and Rewarding Work: Final Report on the Minnesota Family Investment Program. Volume 2: Effects on Children. New York: MDRC. ———. 2002. “Children and Welfare Reform: A View from an Experimental Welfare Program in Minnesota.” Child Development 73(2): 601–19. Gennetian, Lisa A., Pamela Morris, Johannes Bos, and Howard Bloom. 2005. “Constructing Instrumental Variables from Experimental Data to Explore How Treatments Produce Ef-

/

145

Making the Work-Based Safety Net Work Better fects.” In Learning More from Social Experiments: Evolving Analytic Approaches, edited by Howard Bloom. New York: Russell Sage Foundation. Gottfried, Adele E., and Allen Gottfried. 1994. “Role of Maternal and Dual-Earner Employment Status in Children’s Development.” In Redefining Families: Implications for Children’s Development, edited by Adele E. Gottfried and Allen W. Gottfried. New York: Plenum Press. Graber, Judy A., and Jeanne Brooks-Gunn. 1996. “Transitions and Turning Points: Navigating the Passage from Childhood Through Adolescence.” Developmental Psychology 32(4): 768–76. Hamilton, Gayle, Stephen Freedman, Lisa A. Gennetian, Charles Michalopoulos, Johanna Walter, Diana Adams-Ciardullo, Anna Gassman-Pines, Sharon McGroder, Martha Zaslow, Surjeet Ahluwalia, Jennifer Brooks, Electra Small, and Bryan Ricchetti. 2001. How Effective Are Different Welfare-to-Work Approaches? Five-Year Adult and Child Impacts for Eleven Programs. Washington, D.C.: U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation and Administration for Children and Families, and U.S. Department of Education. Harvey, Elizabeth. 1999. “Short-Term and Long-Term Effects of Early Parental Employment on Children of the National Longitudinal Survey of Youth.” Developmental Psychology 35(2): 445–59. Haveman, Robert, and Barbara Wolfe. 1995. “The Determinants of Children’s Attainments: A Review of Methods and Findings.” Journal of Economic Literature 23: 1829–78. Hillman, Stephen B., and Shlomo S. Sawilowsky. 1991. “Maternal Employment and Early Adolescent Substance Use.” Adolescence 26(104): 829–37. Huston, Aletha. 2002. “Reforms and Child Development.” Future of Children 12(1): 59–77. Huston, Aletha C., Greg J. Duncan, Robert Granger, Johannes Bos, Vonnie McLoyd, Rashmita Mistry, Danielle Crosby, Christina Gibson, Katherine Magnuson, Jennifer Romich, and Ana Ventura. 2001. “Work-Based Antipoverty Programs for Parents Can Enhance the School Performance and Social Behavior of Children.” Child Development 72(1): 318–36. Jacob, Brian, and Jens Ludwig. 2006. “The Effects of Family Income on Youth Outcomes: New Evidence from a Housing-Voucher Lottery.” Working paper. Chicago: University of Chicago. Kaushal, Neeraj, Qin Gao, and Jane Waldfogel. 2006. “Welfare Reform and Family Expenditures: How Are Single Mothers Adapting to the New Welfare and Work Regime?” NBER Working Paper No. 12624. Washington, D.C.: National Bureau of Economic Research. Lipsey, Mark, and David B. Wilson. 1996. Tool Kit for Practical Meta-Analysis. Nashville: Vanderbilt University. Ludwig, Jens, and Jeffrey R. Kling. 2007. “Is Crime Contagious?” Journal of Law and Economics 50: 491–518. Ludwig, Jens, and Deborah Phillips. 2007. “The Benefits and Costs of Head Start.” Society for Research in Child Development Social Policy Report 21(3): 3–18. Magnuson, Katherine. 2003. The Effect of Increases in Welfare Mothers’ Education on Their Young Children’s Academic and Behavioral Outcomes: Evidence from the National Evaluation of Welfare-To-Work Strategies Child Outcomes Study. Discussion paper no. 1274–03. Madison: University of Wisconsin, Institute for Research on Poverty.

146

/

Parental Pathways to Self-Sufficiency Mayer, Susan E. 1997. What Money Can’t Buy: Family Income and Children’s Life Chances. Cambridge, Mass.: Harvard University Press. McCall, Robert B. 1981. “Nature-Nurture and the Two Realms of Development: A Proposed Integration with Respect to Mental Development.” Child Development 52(1): 1–12. McGroder, Sharon M., Martha J. Zaslow, Kristin A. Moore, and Suzanne M. LeMenestrel. 2000. The National Evaluation of Welfare-to-Work Strategies: Impacts on Young Children and Their Families Two Years After Enrollment. Findings from the Child Outcomes Study. Washington, D.C.: U.S. Department of Health and Human Services. McLoyd, Vonnie C. 1990. “The Impact of Economic Hardship on Black Families and Children: Psychological Distress, Parenting, and Socioemotional Development.” Child Development 61(2): 311–46. ———. 1997. “The Impact of Poverty and Low Socioeconomic Status on the Socioemotional Functioning of African American Children and Adolescents: Mediating Effects.” In Social and Emotional Adjustment and Family Relations in Ethnic Minority Families, edited by Ronald D. Taylor and Margaret C. Wang. Mahwah, N.J.: Lawrence Erlbaum. ———. 1998. “Children in Poverty, Development, Public Policy, and Practice.” In Handbook of Child Psychology, edited by William Damon. Volume 4: Child Psychology in Practice, edited by Irving Sigel and K. Ann Renninger. 5th ed. New York: Wiley. McLoyd, Vonnie C., Nikki L. Aikens, and Linda M. Burton. 2006. “Childhood Poverty, Policy, and Practice.” In Handbook of Child Psychology, edited by William Damon and Richard W. Lerner. Volume 4: Child Psychology in Practice, edited by K. Ann Renninger and Irving Sigel. 6th ed. New York: Wiley. McLoyd, Vonnie C., Tony Jayaratne, Rosario Ceballo, and Julio Borquez. 1994. “Unemployment and Work Interruption Among African American Single Mothers: Effects on Parenting and Adolescent Socioemotional Functioning.” Child Development 65(2): 562–89. Michalopoulos, Charles, Doug Tattrie, Cynthia Miller, Philip K. Robins, Pamela Morris, David Gyarmati, Cindy Redcross, Kelly Foley, and Reuben Ford. 2002. Making Work Pay: Final Report on the Self-Sufficiency Project for Long-Term Welfare Recipients. New York: MDRC. Miller, Cynthia, Virginia Knox, Martey Dodoo, Jo Anna Hunter, Cindy Redcross. 2000. MFIP: Rewarding Work and Reforming Welfare. Final Report on the Minnesota Family Investment Program. Volume 1: Effects on Adults. New York: MDRC. Morris, Pamela, Greg J. Duncan, and Elizabeth Clark-Kauffman. 2005. “Child Well-Being in an Era of Welfare Reform: The Sensitivity of Transitions in Development to Policy Change.” Developmental Psychology 41(6): 919–32. Morris, Pamela, Greg Duncan, and Christopher Rodrigues. 2006. Does Money Really Matter? Estimating Impacts of Family Income on Children’s Achievement with Data from Social Policy Experiments. Unpublished manuscript. New York: MDRC. Morris, Pamela, and Lisa Gennetian. 2002. Identifying Effects of Income on Children’s Development: Integrating an Instrumental Variables Analytic Method with an Experimental Design. New York: MDRC. Morris, Pamela, and Richard Hendra. 2009. “Losing the Safety Net: How Welfare Time Limits Affect Families and Children?” Developmental Psychology 45(2): 383–400. Morris, Pamela, and Charles Michalopoulos. 2000. The Self Sufficiency Project at 36 Months: Effects on Children of a Program that Increased Employment and Income. New York: MDRC.

/

147

Making the Work-Based Safety Net Work Better Muller, Chandra. 1995. “Maternal Employment, Parent Involvement, and Mathematics Achievement.” Journal of Marriage and the Family 57(1): 85–100. National Institute of Child Health and Human Development, Early Child Care Research Network. 1997. “Poverty And Patterns of Child Care.” In Consequences of Growing Up Poor, edited by Jeanne Brooks-Gunn and Greg Duncan. New York: Russell Sage Foundation. ———. 2006. “Duration and Developmental Timing of Poverty and Children’s Cognitive and Social Development from Birth Through Third Grade.” Child Development 76(4): 795–810. National Institute of Child Health and Human Development, Early Child Care Research Network, and Greg J. Duncan. 2003. “Modeling the Impacts of Child–Care Quality on Children’s Preschool Cognitive Development.” Child Development 74(5): 1454–75. Paulson, Sharon E. 1994. “Relations of Parenting Style and Paternal Involvement with Ninth Grade Students’ Achievement.” Journal of Early Adolescence 14(2): 250–67. Richards, Maryse H., and Elena Duckett. 1994 “The Relationship of Maternal Employment to Early Adolescent Daily Experience with and Without Parents.” Child Development 65(1): 225–36. Salkind, Neil J., and Ron Haskins. 1982. “Negative Income Tax: The Impact on Children from Low–Income Families.” Journal of Family Issues 3(2): 165–80. Shonkoff, Jack P., and Deborah A. Phillips. 2000. From Neurons to Neighborhoods: The Science of Early Childhood Development. Washington, D.C.: National Academy Press. Smith, Judith, Jeanne Brooks–Gunn, and Pamela K. Klebanov. 1997. “Consequences of Living in Poverty for Young Children’s Cognitive and Verbal Ability and Early School Achievement” In Consequences of Growing Up Poor, edited by Greg Duncan and Jeanne Brooks–Gunn. New York: Russell Sage Foundation. Smolensky, Eugene, and Jennifer A. Gootman. 2001. Working Families and Growing Kids. Washington, D.C.: National Academies Press. Sugland, Barbara W., Martha J. Zaslow, Judith R. Smith, Jeanne Brooks–Gunn, Kristin A. Moore, Connie Blumenthal, Terri Griffin, and Robert Bradley. 1995. “The Early Childhood HOME Inventory and HOME Short Form in Differing Socio–Cultural Groups: Are There Differences in Underlying Structure, Internal Consistency of Subscales, and Patterns of Prediction?” Journal of Family Issues 16(5): 632–63. Vandell, Deborah L., and Janaki Ramanan. 1992. “Effects of Early and Recent Maternal Employment on Children from Low–Income Families.” Child Development 63(4): 938–49. Vander Ven, Thomas M., Francis T. Cullen, Mark A. Carrozza, and John Paul Wright. 2001. “Home Alone: The Impact of Maternal Employment on Delinquency.” Social Problems 48(2): 236–57. Waldfogel, Jane. 2006. What Children Need. Cambridge Mass.: Harvard University Press. Yeung, Wei–Jun, Miriam R. Linver, and Jeanne Brooks–Gunn. 2002. “How Money Matters for Young Children’s Development: Parental Investment and Family Processes.” Child Development 73(6): 1861–79. Zaslow, Martha J., and Carol A. Emig. 1997. “When Low–Income Mothers Go to Work: Implications for Children.” Future of Children 71(2): 110–15.

148

/

Chapter 5 School Reforms and Improved Life Outcomes for Disadvantaged Children David N. Figlio

I

n this chapter I describe the treatment of disadvantaged families in the United States education system today, and present an economics perspective regarding the likely consequences of education reforms, both popular and lesstested, in terms of improving the life outcomes of disadvantaged children. In this chapter I concentrate on the potential education reforms where economics offers the most insight—policies concerning school spending and class size; teacher quality and teacher compensation; and other market-based policies such as school choice and school accountability. In each case I describe the rationales for the proposed policies and discuss the distributional consequences of the policies, especially the potential effects on disadvantaged children, should the policies be universally adopted. Thinking about education reform is so important for promoting self-sufficiency because of the large and pronounced relationship between educational attainment and self-sufficiency in the United States, and because of the strong intergenerational correlations in educational attainment. According to census data, over 25 percent of all adults with less than high school education lived in poverty in 2000, whereas only 2.5 percent of college graduates had incomes below the poverty level. The relationship between educational attainment and adult earnings in the United States has only increased over time. In 1979, college-educated adults had 75 percent higher earnings than did high school graduates, and by 2003 college-educated adults averaged 2.3 times the annual earnings of high school graduates (Rouse and Barrow 2006). High school dropouts have fared even worse: Cecilia Rouse and Lisa Barrow (2006) report that high school dropouts work in the lowest-paid occupational groups at more than twice the rate of high school graduates. While the precise degree to which educational attainment influ-

/

149

Making the Work-Based Safety Net Work Better

ences subsequent economic well-being is still under debate, the evidence is clear that those who remain in school longer tend to earn considerably greater incomes as adults. A consensus has emerged that the rate of return to an extra year of schooling is at least 10 percent (Carneiro and Heckman 2003). It is also evident that by any metric, children from lower socioeconomic status (SES) homes tend to have lower educational attainment. For example, around 95 percent of children from the highest quartile of the family SES distribution ultimately graduate from high school, but according to the National Education Longitudinal Study of 1998, only about 66 percent of those from the lowest SES quartile graduate. Although it is not certain exactly how much of these differences are due to the causal pathways of parental SES and how much are due to unobserved correlations between parents and children, the weight of the evidence suggests that parental SES has a strong influence on children’s outcomes. Regardless of whether or not these relationships are causal or purely correlative in nature, they present significant challenges for those concerned with maximizing the life chances of children from low-income families. What is clear is that children from low-income families begin school at a considerable disadvantage. In the kindergarten class of 1998, 52 percent of children of mothers who are high school dropouts scored in the bottom quartile of the kindergarten reading distribution, as compared with only 8 percent of children of college-graduate mothers (West, Denton, and Germino-Hausken 2000). At the other end of the spectrum, only 6 percent of children of high school–dropout mothers scored in the top quartile of the reading distribution whereas 46 percent of children of college-graduate mothers scored in the top quartile for reading. Similar patterns exist in other areas as well: 53 percent of children of high school– dropout mothers scored in the bottom quartile in mathematics and 52 percent of these children scored in the bottom quartile in general knowledge. These rates are nearly identical if one looks at children of mothers who had ever received welfare. The large gaps in kindergarten readiness that exist between those from relatively advantaged versus relatively disadvantaged families suggest an important role for K-through-twelve education policy in the United States. Students from disadvantaged households enjoy considerably lower educational attainment, even when taking their performance in school into account, than do their counterparts from relatively advantaged families. Putting the pieces of the puzzle together, it becomes clear that students from relatively disadvantaged backgrounds are at increased risk of failing to become self-sufficient as adults. Education policy over the past four decades has sought to break this link between lack of self-sufficiency of adults and a lack of self-sufficiency of their children. But despite the oft-stated goals of fostering performance and self-sufficiency among the more marginal students, it is apparent that commonly proposed forms of remediation of the outcomes of children of disadvantaged families do little to improve the overall intergenerational transmission of human capital. By the end of fifth grade, the gap between advantaged and disadvantaged children has hardly changed from that in kindergarten in the Early Childhood Longitudinal Program–Kindergarten (ECLS-K) database. Sixty-six percent of children of high

150

/

School Reforms and Improved Life Outcomes

school–dropout mothers are in the bottom third of the reading distribution and 62 percent are in the bottom third of the math distribution (Princiotta, Flanagan, and Germino-Hausken 2006). These are similar figures to those for children currently in poverty (61 and 57 percent, respectively) and those who have been in poverty consistently since kindergarten (67 and 66 percent, respectively). They differ dramatically from those of children of college-educated mothers: only 10 percent of these children score in the bottom third of the fifth-grade reading distribution and 13 percent score in the bottom third of the fifth-grade math distribution. These gaps continue to grow as children progress through school. Among high school sophomores in 2002, children from the top SES quartile averaged 12.5 more questions correct (out of approximately forty total questions) on the itemresponse theory mathematics test administered by the Education Longitudinal Study of 2000 than did those from the bottom SES quartile (Cahalan et al. 2006). Put into perspective, this means that while 71 percent of high SES children were proficient at level three (“simple problem solving, requiring the understanding of low level mathematical concepts”), only 25 percent of low SES children were proficient. Given the strong policy interest in improving the performance of low-income students, it is striking that these gaps in tenth-grade performance have remained virtually unchanged for a generation. The comparable mathematics gap in 1990, taken from the National Education Longitudinal Study of 1988, was 13.2 questions, and the gap in 1980, taken from the High School and Beyond Study, was 12.5 questions. The same gaps and temporal stability are evident with regard to reading, as well as in the National Assessment of Educational Progress (NAEP), otherwise known as the “Nation’s Report Card.” The average mathematics NAEP score difference in twelfth grade between children of college graduates and children of high school dropouts was thirty points in 2004, down slightly from thirtyseven points in 1978. For reading, the twelfth-grade gap was thirty-nine points in both 1980 (the first available year of data) and 2004. At least as important is that these constant gaps do not mask substantial improvements over time in performance across the spectrum: apples-to-apples comparisons of gains over the last quarter century have been small at best. That these gaps in achievement have remained so constant over time in an era of active financial and pedagogical attempts to increase equity in educational opportunity indicates that there are no easy answers to the question of how to best educate children from disadvantaged backgrounds. One major explanation for these large and persistent gaps in educational outcomes is that schools provide just some of the many inputs of the education production function. If parental income or education matters above and beyond the improved schooling that is purchased or demanded by relatively advantaged families, then one might fully expect children from low-income families to lag behind, despite the efforts of schools. The same is true if children from disadvantaged families tend to be of lower innate ability, though the evidence on this point is debatable. Much less controversial is the assertion that high-income and educated parents offer their children a considerably more robust leg up than disadvantaged children before

/

151

Making the Work-Based Safety Net Work Better

formal schooling begins—evidenced by the large gaps in kindergarten readiness—and that they are more capable of securing superior schooling arrangements that maintain or this advantage as the children enter adulthood. Derek Neal and William Johnson (1996) suggest that low-income families provide lowerquality parental inputs for their children’s human capital. And income appears to matter as well: Gordon Dahl and Lance Lochner (2006), among others, demonstrate that even transitory income shocks can affect student performance while in school, and Philip Oreopoulos, Marianne Page, and Ann Huff Stevens (2005) and others show that parental income shocks can strongly affect educational attainment. It is reasonable to expect that schools provide equality of educational opportunities for all children, but it is unreasonable to expect equality of educational outcomes. That said, there is reason to believe that the educational outcomes of disadvantaged children could be substantially improved by educational policies and practices, and therefore it should be possible to dramatically increase the chances that disadvantaged children attain self-sufficiency as adults. This chapter makes the case that the key to improving educational outcomes for disadvantaged children, and hence improving their later self-sufficiency chances, involves improving teacher quality. This could possibly take place through increasing the teaching quality and work effort of existing teachers, or recruiting higher-quality new teachers. Identifying quality teachers is difficult, however, as is recruitment and retention of these individuals. I discuss a set of potential policy options that may work to improve teacher quality for disadvantaged students, and describe the promises and pitfalls of these potential solutions.

SCHOOL SPENDING AND DISADVANTAGED CHILDREN For the last three decades, states across the country have altered their school finance systems in order to reduce the relationship between family wealth and the quality of schooling that a child has available. These school finance reforms have largely been motivated by judicial interpretations of state constitutional language requiring some notion of “adequacy” of educational provision, or “equality” of educational opportunity, though they have also been promoted in some instances by legislators or state executives interested in workforce development. The rationale behind these school finance reforms is that in a system in which the local property tax is the major source of school district revenues and individuals are sorted into neighborhoods on the basis of income or wealth, disadvantaged students will tend to attend schools that have considerably fewer resources and that are, as a consequence, of substantially lower quality than the schools that more advantaged students can attend. By either lessening the importance of the property tax altogether as a mechanism for funding schools (and instead, increasing the reliance on sales or income taxes), utilizing state sources of revenue as a redistributive tool, or maintaining the primacy of the property tax as a school finance mechanism but capturing some of wealthy districts’ property tax revenues for re-

152

/

School Reforms and Improved Life Outcomes

allocation to poorer districts, school finance reforms seek to level the playing field by reducing the link between family background and spending.

Spending Alone Will Not Significantly Improve Disadvantaged Children’s Outcomes Today, partially as a result of these school finance reforms (Murray, Evans, and Schwab 1998; Card and Payne 2002; Springer, Liu, and Guthrie 2005), there exists very little relationship between family income and school district spending in the United States. In fact, the 20 percent of American school districts serving the leastadvantaged populations average 8 percent more expenditures per pupil than do the 20 percent of American school districts serving the most advantaged populations (Rouse and Barrow 2006). Even though some authors have found limited evidence of increased equality in outcomes (see, for instance, Card and Payne 2002), in general these school finance reforms have not been coupled with dramatic improvements in student achievement (Downes 1992; Clark 2003; Roy 2004; Hoxby 2000). Quasi-experimental approaches to investigating the relationship between school spending and student outcomes more often lead to positive estimated effects of spending on measured test scores. For instance, Jonathan Guryan (2003) finds that increasing school spending by $1,000 per pupil in Massachusetts led to test score gains of one-third of a standard deviation or more. However, the fact that school spending has been considerably equalized in the United States but few improvements have occurred in the performance of disadvantaged children over the same time period has led some researchers to conclude that increasing school spending per se will lead to little or no improvements in the desired educational outcomes (Hanushek 1996). There are many possible reasons why increased spending in disadvantaged communities would not necessarily lead to dramatically improved outcomes. One possible reason is that the costs of educating children from different backgrounds may themselves vary considerably. For one, disadvantaged children tend to attend school in districts with older infrastructure that requires higher maintenance costs. In addition, the rates of special education in disadvantaged communities far outstrip those in more advantaged school districts. As students with special educational needs require more costly teachers and school environments, dollars spent per pupil is not the most appropriate comparison across school districts with differing degrees of advantage. Moreover, low-income school districts tend to have more difficulty recruiting and retaining higher-quality teachers (Lankford et al. 2002; Clotfelter, Ladd, and Vigdor 2006), indicating that raw dollar counts do not translate well into differences in school quality. And there is no guarantee that increasing school spending, even in a dramatic fashion, will lead to vastly improved student outcomes and likely improvements in long-run self-sufficiency. Absent strong market-disciplining mechanisms, schools may face little incentive to efficiently deploy increased resources. Increasing teacher salaries may improve potential teacher quality along some measured

/

153

Making the Work-Based Safety Net Work Better

dimensions (Figlio 1997; Bacolod 2007), but schools might not hire the best candidates from their pool of applicants (Ballou 1996; Ballou and Podgursky 1997). And schools may spend their increased resources on inputs that are politically popular or that are widely believed to improve outcomes but that actually have more limited effects on student learning. Schools’ inability to identify the most effective uses of their resources blunts their capacity to improve student outcomes in a dramatic fashion.

Targeting Spending to Reduced Class Size Since merely increasing spending on schools serving disadvantaged children seems to not greatly improve student outcomes, state policymakers have frequently proposed targeting spending to specific purposes. The most popular of these targeted spending proposals involves reducing class sizes for all students in the state, and not merely the least advantaged. Today, two-thirds of all states have class size caps, and some states—including California, Florida, North Carolina, and Texas—have in place policies to actively reduce class sizes in the elementary grades. The rationale behind reducing class size is clear: teachers who are responsible for fewer students at a time can devote more individual attention to students and have a wider array of potential teaching approaches in their toolkit than would teachers who must teach larger classes. And as Edward Lazear (2001) points out, smaller classes can reduce the degree of classroom disruption, which David Figlio (2007) shows to have strong negative effects on student learning in the classroom. The findings of the negative consequences of disruption are particularly pronounced in classrooms serving disadvantaged students. But there are also numerous reasons why class size reductions may not lead to improvements in student outcomes. One possibility is that teachers may not be equipped to take advantage of the smaller classes, and may not alter their behavior in response to the reduced class size. Another possibility is that the critical value for improved outcomes due to class size reductions may be below the range of class sizes that are in play. If classroom disruptions cause the same problems in a class of fifteen as in a class of twenty-five, then there may not be appreciable benefits of class size reductions of ten students. Likewise, it may be the case that teachers require extremely small class sizes before they can effectively alter their classroom management and instructional styles. Therefore, measuring the benefits of class size reductions is an empirical question. Early work on this subject tended to focus on student-teacher ratios as a proxy for class size. However, although student-teacher ratios are certainly correlated with class size, this correlation can be quite low. Michael Boozer and Rouse (2001), for instance, find that the lower student-teacher ratios in schools that serve disadvantaged populations are due to increased rates of special education and limited English proficiency. That said, David Card and Alan Krueger (1992) show that states that reduced their student-teacher ratios tended to have increases in adult earnings. This finding provides some evidence of a long-run benefit to reduced

154

/

School Reforms and Improved Life Outcomes

class sizes, at least as measured by the weak proxy of smaller student-teacher ratios. The most compelling evidence regarding class size reductions, however, comes from Tennessee’s Student-Teacher Achievement Ratio (Project STAR) experiment, in which students in 1985 to 1986 were randomly assigned to either a small class with thirteen to seventeen students, a regular-sized class of twenty-two to twentyfive students, or a regular-sized class with a full-time teacher’s aide. To reduce the likelihood of nonrandom selection, teachers were also randomized (within a school) to the different types of classes. Students were intended to remain in the same randomly assigned class type through third grade. In practice, however, a large number of students entered the project after kindergarten and a significant fraction left the study because of changing schools, grade retention, or grade skipping, and about 10 percent of participants changed class types in a nonrandom manner (Whitmore Schanzenbach 2006). Much of the research on Project STAR has focused on attempting to take advantage of the study design while dealing as well as possible with the inference problems associated with the real-life implementation of the experiment. These problems have led most researchers to focus on intent-to-treat approaches of estimating the effects of class size reductions— that is, focusing on potential recipients of a treatment rather than on those who actually avail themselves of it. Problems aside, the weight of the evidence from Project STAR suggests that students across the distribution benefited from smaller class sizes (Finn and Achilles 1990; Word et al. 1990; Krueger 1999; Krueger and Whitmore 2001; Nye, Hedges, and Konstantopoulos 2002) in terms of test score improvements through third grade. The estimated benefits were particularly large for African American students and students from disadvantaged backgrounds, indicating that class size reductions have the potential for substantially improving the outcomes of children from low-income families. Follow-up studies by Krueger and Diane Whitmore (2001, 2002) show that students assigned to smaller classes experienced persistent increases in test scores and that black students assigned to smaller classes in the early grades showed increased rates of college entrance exam test-taking and scores. Krueger and Whitmore conclude that smaller classes could have large effects in reducing the magnitude of the persistent black-white test score gap, and it follows that significant reductions in test score gaps by economic background could occur as well. Class size reductions may present a serious scalability problem. If a state or a large school district reduces its class sizes, it requires a substantial increase in the number of teachers, which may necessitate the hiring of large numbers of inexperienced or potentially inadequately trained teachers; this could account for the disappointing results of class size reductions in California (Jepsen and Rivkin 2001; Bohrnstedt and Stecher 2002), where little improvement in test scores has been observed resulting from smaller class sizes. This increase in the demand across the board for new teachers may actually have negative distributional consequences for disadvantaged students. Since schools serving relatively advantaged students will also have an increased de-

/

155

Making the Work-Based Safety Net Work Better

mand for teachers, higher-quality or more qualified teachers currently teaching in low-income schools could move to more advantaged schools. Given the tendency for this phenomenon to happen absent a class size reduction (Boyd et al. 2006; Lankford et al. 2002; Clotfelter, Ladd, and Vigdor 2006; Scafidi, Sjoquist, and Stinebrickner 2007; Hanushek et al. 2007; Reed, Reuben, and Barbour 2006), there is ample reason to expect that an additional class size reduction would only exacerbate this pattern. This would leave students in low-income schools with a larger number of less-experienced and likely lower-quality teachers. To the extent to which this is the case, class size reductions in general equilibrium have the potential to worsen outcomes for disadvantaged children. Evidence that teacher quality almost surely has larger effects than class size even in the controlled setting of the Tennessee Project STAR experiment (Hanushek 1999) furthers this concern. A class size reduction that is specifically targeted toward disadvantaged students and schools is therefore more likely to improve the outcomes of students from low-income families than an across-the-board reduction.

THE FUNDAMENTAL IMPORTANCE OF TEACHERS The distributional consequences of school reforms such as class size reductions depend crucially on the nature of the teaching force—specifically, the variation in teacher quality across potential teachers. Most readers’ personal experiences as students or parents likely lead to the realization that there exists massive variation in teacher quality in the United States, and even within given schools. These suspicions are confirmed by data: Steven Rivkin, Hanushek, and John Kain (2005) estimate that a one-standard-deviation increase in teacher quality, identified using so-called “value added” models of student outcomes, is associated with student achievement gains of four national percentile ranks within a given year—a remarkable achievement. And these estimates are likely a lower bound of the true variation in teacher effectiveness, as the data employed in that paper are reported at the school-by-grade level. To the extent to which there exists variation in teacher quality within a grade level in a school, studies that assume that all teachers within a grade are equally effective will likely understate the effects of teachers on student performance, and indeed, Jonah Rockoff’s (2004) research linking students to teachers indicates an even larger effect of teacher quality on student outcomes. In earlier work, Eric Hanushek (1992) presents evidence that for low-income minority students—precisely the set of marginal students most at risk of not being self-sufficient as adults—very high quality teachers could have the marginal effect of a full additional year of learning, relative to very low quality teachers. Figlio and Maurice Lucas (2004) find that teachers within a school have very different standards for evaluating their students, and these standards translate into large differences in student outcomes. It is clear that very large variation exists in teacher quality, even within a single grade level in a single school. Given the large variation in teacher quality present in the United States and

156

/

School Reforms and Improved Life Outcomes

within individual districts and schools, major improvements in student outcomes could be realized if schools were to identify the most successful teachers and deploy these teachers in the settings in which they would be most productive. But a crucial problem in identifying and deploying successful teachers is that it is very difficult to identify just who will become a successful teacher. Certain observable factors, such as teacher experience, are associated with student outcomes, but blunt instruments such as teacher education levels (including subject-matter expertise as well as the quantity of education), teacher test scores, form of teacher certification, and teacher experience levels together only explain a minuscule portion of the variation in measured teacher quality. If the information at school districts’ disposal at the time of teacher hiring is weak, it only stands to reason that school districts will have difficulty identifying the potential teachers who have the most promise as educators. Deploying high-quality teachers, once their quality is realized, to the settings in which they would be most productive presents its own set of challenges. For one, the identification of individual high-quality teachers in practice is more difficult than the measurement of variation in teacher quality, and recent evidence (Rothstein 2007a) raises questions about whether even the simpler task is as possible as researchers have proposed. But even if one could rank-order teacher quality with certainty, schools and districts face institutional hurdles to strategically matching teachers to classrooms. Most notably, collective bargaining rules typically provide teachers with considerable discretion as to where they will teach, and as mentioned previously, better-qualified teachers tend to move away from schools serving disadvantaged populations when given the opportunity. And salary structures that provide tenure to teachers before their true productivity is known and that pay teachers on the basis of their education and experience levels rather than on productivity make it very difficult to fire or counsel out poorly performing teachers. Given the current structure of teacher compensation nearly universally prevalent in the United States, schools and districts have very limited opportunities to act upon information about differential teacher quality.

Changing the Structure of Teacher Compensation One method through which schools, districts, and states may attempt to improve the quality of teachers across the board is via increased teacher salaries. Increasing teacher salaries in general is often motivated by the goal of improving the quality of the new-teacher pool and reducing attrition of highly qualified teachers from the existing teacher pool (Corcoran, Evans and Schwab 2004; Hoxby and Leigh 2004). However, there exists substantial evidence (see, for instance, Rivkin, Hanushek, and Kain 1999; Loeb and Page 2000) that raising salaries is unlikely to considerably improve the quality of the teaching force, and could actually have the opposite effect, if higher salaries induce fewer poor-quality teachers to leave the profession, as Hanushek has found to be the case in Rochester schools. Nor will offering differential pay for teachers in disadvantaged schools necessarily

/

157

Making the Work-Based Safety Net Work Better

lead to improved teacher quality in those schools: R. Clarke Fowler (2003) presents evidence that one-time signing bonuses for teachers to teach in low-performing schools are unlikely to lead to long-term improvements in teacher quality, as they do not incentivize teachers to remain in the low-performing schools after the signing bonus. Rather than increasing teacher salaries across the board or offering bonuses for teachers to move to schools serving disadvantaged students, states or school districts could instead offer targeted bonuses to keep teachers in high-poverty schools. Charles Clotfelter, Helen Ladd, and Jacob Vigdor (2006) study the effects of just such a policy in North Carolina, where the state awarded an annual bonus of $1,800 to qualified math, science, and special education teachers working in high-poverty schools or schools deemed to be academically failing. Utilizing longitudinal data on teachers, they estimate that these bonuses reduced turnover rates of the targeted teachers—especially the more experienced teachers. Moreover, there are at least two reasons to believe that the sizable turnover reductions are understatements of the potential: First, the state did a poor job of making eligible teachers aware of their eligibility for these bonuses; this advertising problem suggests that a better-administered bonus program could have even larger retention benefits. Second, the retention bonus is still quite small—about 4 percent of a typical teacher’s salary. If teachers are responsive to relatively small retention bonuses, it may be the case that larger bonuses would be even more beneficial. An important caveat, however, is that although teacher experience is correlated with student outcomes, it is only a weak proxy for teacher quality; it remains unknown whether policies such as the North Carolina bonus lead to large improvements in teacher quality in disadvantaged schools. A different mechanism for using teacher compensation to improve the performance of disadvantaged students involves offering differential pay not for teaching in a disadvantaged school but rather for demonstrating a high degree of competence in teaching. Advocates for performance pay for teachers argue that teachers typically receive the same salary regardless of their performance, meaning that other than intrinsic motivations for doing a good job there is little in the way of external motivation to do so. With this in mind, a number of states (one is Florida) and school districts (in Denver, Minneapolis, and Nashville) are employing some notion of merit pay in their teacher compensation policies. A significant argument for teacher pay for performance comes from the “new personnel economics” literature associated with Lazear: productivity is improved when firms pay for outputs rather than for inputs, and the same may be true for schools and teachers. Indeed, the potential gains for schools may be even greater, as it’s particularly hard to identify the best teachers on the way in, and a single salary schedule makes it hard to identify the teachers who will turn out to be more productive. According to theory, pay for performance will tend to attract and retain individuals who are good at that which is incentivized and repel those who aren’t; therefore, even if merit pay doesn’t improve individual teacher performance, it can still improve overall outcomes purely through labor-market factors. On the other hand, there are numerous reasons to be opposed to paying teach-

158

/

School Reforms and Improved Life Outcomes

ers for their measured performance. One major argument is that teachers engage in team production. Although this could be dealt with through the use of teambased performance pay, doing so introduces the potential for a free-rider problem, in which teachers may reduce their efforts because they are compensated on the basis of both their own efforts and the efforts of others. Another concern with teacher performance pay is that teachers’ performance has multiple dimensions, and rewarding on one dimension could cause shirking on other dimensions. Still another concern with the use of teacher merit pay is that teacher performance is arguably more difficult to measure than performance in many other professions— test scores are indicators rather than actual outputs of production. There exists considerable research on the factors underlying schools’ decisions to implement teacher merit pay plans (Ballou and Podgursky 1997; Ballou 2001) and the stability of these plans (Johnson 1986; Murnane and Cohen 1986; Hatry, Greiner, and Ashford 1994), but little research has been conducted to date on the efficacy of merit pay in the United States. Clotfelter and Ladd (1996) and Ladd (1999) find that school-based performance incentives appear to improve student performance, and Victor Lavy (2002) shows that school-based incentive programs in Israel are also associated with higher test scores. Paul Glewwe, Ilias, and Kremer, (2004), on the other hand, suggest that school-based programs in rural Kenya have short-term apparent benefits but no long-term benefits, suggesting that participating teachers may have attempted to “game the system.” It is a stretch to apply lessons learned from places such as rural Kenya to the United States, but the importance of work such as Glewwe et al.’s underscores how little is known about the potential effectiveness of teacher merit pay systems. Still less is known about the potential benefits of individual-based teacher merit pay. Figlio and Lawrence Kenny (2007) match test score data on a low-stakes test for schools (the National Education Longitudinal Study) to merit pay questions and find evidence that schools offering judiciously administered performance bonuses (that is, large bonuses to relatively few teachers) have larger test score gains than schools that do not offer performance pay. Moreover, they find that the positive estimated effects of performance pay are strongest in relatively disadvantaged schools, where parental monitoring of teachers may be lower. Their evidence indicates that performance pay has the potential to significantly benefit disadvantaged students, though these conclusions should be treated with caution because of the cross-sectional nature of their evidence. Thomas Dee and Benjamin Keys (2004), employing data from the Tennessee Project STAR experiment, present evidence indicating that students assigned to teachers who were participants in a system that resembles performance pay in some dimensions experienced gains in mathematics though not in reading test scores. More research is needed, but the evidence to date suggests that teacher merit pay has the potential to improve student outcomes in disadvantaged schools. Randomized experiments such as those funded by the U.S. Department of Education that began in fall 2007 will provide valuable evidence on the causal effects of individual teacher incentives, as schools are being randomized into different groups based on the presence of teacher performance pay. In the meantime, experimental evidence from a

/

159

Making the Work-Based Safety Net Work Better

World Bank demonstration project in rural India (Muralidharan and Sundararaman 2007) also finds evidence of positive performance effects of both individual and collective teacher merit pay on student test scores. Lavy (2004) finds that individual teacher performance tournaments in Israel also led to improved student outcomes. Thus, existing evidence indicates that teacher performance pay has the potential to be successful in improving school productivity. It is important to note, moreover, that these performance increases are likely to be substantial understatements of the full effects of teacher merit pay systems on student outcomes. The evidence presented on individual teacher incentives concentrates solely on short-term incentives, which can only be expected to boost effort among existing teachers. But if teacher performance pay programs in the long run alter the nature of individuals selecting into teaching in favor of potential teachers who are particularly skilled at improving performance along the measured dimensions, then it stands to reason that teacher performance pay would have the long-run consequence of boosting student outcomes through the overall improvement of the teaching force (at least as measured by the factors upon which the teacher merit pay system are based). In order for this long-term change in teacher quality to be realized, however, certain features of the performance pay system should be present. For one, the performance pay system should be perceived to be permanent rather than transitory. The history of performance pay systems in the United States is that they tend to be short-lived. But systematic change in the nature of who selects into teaching on the basis of the performance pay system would crucially depend on potential teachers’ perceptions about the long-term availability of performance pay. Second, the performance pay system should be perceived as being fair to all teachers. Teacher merit pay systems should take into account differences in student background, so that the students assigned to a teacher—the teacher’s “draw”—does not determine his or her likelihood of receiving performance bonuses. This is particularly difficult to do: It was recently revealed that Hillsborough County’s (Tampa, Florida) performance pay system, which was intended to take student background explicitly into account, still in 2007 dramatically disproportionately distributed performance pay to teachers in higher-income communities. This fact could be due to the long-term movement of more qualified teachers to advantaged schools, but it also indicates that system design must be done extremely carefully. Third, performance pay systems may require considerably higher average teacher salaries in order to ensure teachers against the increased risk inherent in these systems. Therefore, although performance pay systems present many intriguing opportunities, they must be implemented very carefully in order to increase their likelihood of success.

Increasing Alternative Opportunities for Teaching For education reforms to have large-scale effects on student outcomes, it is likely that there must be an expansion of the potential teaching pool. That is, education

160

/

School Reforms and Improved Life Outcomes

reforms must influence not only the effort of existing teachers but also the composition of the set of people interested in teaching. One way to affect the pool of potential teachers is to raise salaries across the board for teachers, but as mentioned earlier, this could have deleterious effects on the realized composition of the teaching force. A different way in which the pool of potential teachers could be influenced involves changing the price of becoming a teacher. Currently, most teachers enter the teaching force through traditional means, which entails committing to several years of formal education training, along with major financial commitments that are not easily transferrable to other fields of employment. This training may have value for classroom management, assessment, and instructional practice (though the evidence on specific credentials and training suggests that there is a limited effect on student outcomes). But it is also very costly for potential teachers, for whom the training would have little value in other fields besides education. Therefore, the high costs of obtaining teacher credentialing can present a large barrier to entry for potential teachers, and numerous scholars have suggested that these entry barriers are particularly effective at preventing some of the most talented Americans from trying their hand at teaching. Some jurisdictions have in recent years expanded alternative pathways to teacher certification, bypassing the traditional teacher training routes. Early evidence on alternative pathways suggests that these alternative approaches to staffing schools might not benefit disadvantaged students. Donald Boyd et al. (2006) show that teachers in New York who entered teaching through alternative pathways tend to benefit white and nonpoor students, but there is little evidence that these benefits are enjoyed by minority and low-income families. It is highly possible, however, that a self-selection mechanism is at play here, where highly capable alternatively certified teachers might tend to work at relatively advantaged schools and more marginal alternatively certified teachers might end up in schools serving a disadvantaged population. Probably the most prominent program offering alternative pathways to teaching is the Teach For America (TFA) program, which is specifically targeted to lowincome schools. TFA recruits college seniors and recent graduates to teach for a minimum of two years in disadvantaged schools, and specifically targets students from selective colleges and other standout students who have typically not received formal education training. Participants in the TFA program have only five weeks of formal training in teaching practice, curriculum development, and classroom management, and lack virtually all of the formal pedagogical and practice training that traditionally certified teachers have prior to entering the classroom. An experimental random-assignment study of TFA trainees’ effectiveness in Baltimore, Chicago, Houston, Los Angeles, New Orleans, and the Mississippi Delta shows that even though TFA participants are much less likely to have substantial formal educational training than are traditionally certified teachers, they are dramatically more likely to have graduated from very selective undergraduate institutions and to have had a liberal arts training in college (Decker, Mayer, and Glazerman 2004). TFA participants also generated modest improvements in

/

161

Making the Work-Based Safety Net Work Better

mathematics performance and comparable gains in reading performance when compared to traditionally certified teachers. The results of this evaluation indicate that having a TFA participant for a teacher did not harm and potentially helped student performance, relative to having a traditionally certified teacher. To the extent to which pedagogical training improves student outcomes (as DarlingHammond et al. 2006 find to be the case in a study of Houston students and teachers) these results indicate that the benefits from having a motivated teacher with high levels of individual skills counteracts the costs of having a teacher without formal teacher preparation. Moving forward, it appears that recruiting highly skilled and motivated graduates from selective colleges who don’t have formal teacher preparation could pay considerable dividends, at least if doing so is coupled with more training than the five-week crash course currently offered to TFA participants. The nature of this additional training is not obvious, however, as the evidence of effectiveness of professional development for existing teachers is rather uninspiring (Harris and Sass 2007). Moreover, as Linda Darling-Hammond et al. (2006) point out, TFA participants tend to leave teaching within three years of beginning. That said, the evidence indicates that widening the channels into teaching for potentially very highly competent people is likely to help and unlikely to hurt student performance, and experimentation to help encourage these individuals to stay in teaching as well as to provide additional pedagogical, curriculum development, and classroom management training to alternatively certified teachers could bear very substantial fruit.

OTHER MARKET-BASED POLICIES TO IMPROVE SCHOOL PERFORMANCE An alternative approach that might be used to improve the performance of schools serving disadvantaged students involves market-based policies. These policies broadly fall into two categories. One form of market-based mechanism is in effect an information-based approach, involving alignment of school curriculum with performance standards for students, and holding schools accountable for student outcomes. In this type of approach, schools whose students perform particularly well or particularly poorly relative to objective standards would be publicly called out, and possibly other rewards and sanctions in addition to accolades or stigma could accompany these pronouncements of school quality. A second thrust of market-based policies involves increasing school autonomy and allowing families to choose amongst competing schools. The notion behind this choice-based mechanism would be that schools would risk losing their students (and hence, teachers and administrators may risk losing their jobs) when their performance is benchmarked against those of other schools. The idea behind both of these forms of market intervention is that schools would face new incentives (from either accountability, increased competition, or both) to pay close attention to both the ways their existing teachers educate their students and the quality of their existing teachers themselves. Schools may exper-

162

/

School Reforms and Improved Life Outcomes

iment more with new instructional policies and practices, and may be more likely to make personnel decisions with student performance in mind. One potential consequence of this increased awareness could be that schools would change the mix of teachers they hire and retain, without formal central government intervention. This is particularly likely to happen in the long run if, as with teacher incentive plans, the market-based mechanisms are perceived to be around for the long haul. Both forms of market-based mechanisms are motivated by the economists’ model of the principal-agent problem. In such a model, school administrators and teachers may underperform because the state policymakers do not have a good means of monitoring them. It follows that student achievement would improve if state policymakers could monitor the teachers and school administrators more effectively. School accountability systems serve as a mechanism for counteracting this principal-agent problem. By assessing schools against a common metric, policymakers generate independent information about the performance of schools and school districts. This common yardstick then allows policymakers to compare each school’s performance to that of another school or to an external standard. By measuring, reporting, and, in many cases, attaching positive consequences to strong school performance and negative consequences to weak performance, policymakers provide incentives for schools and school districts to focus attention on what is being measured and possibly to change the way they deliver education. Choice-based mechanisms could have similar consequences.

School-Based Accountability School accountability typically operates within the traditional public school system and relies heavily on student testing (Clotfelter and Ladd 1996; Hanushek and Raymond 2003). Most emblematic is the federal No Child Left Behind Act (NCLB), which became law in 2002. NCLB requires states to test students in reading and mathematics in grades three through eight, as well as in one high school grade. In addition, it requires states to assess schools on the basis of whether their students (both in the aggregate and by subgroup) are making adequate yearly progress (AYP) toward the ultimate goal of 100 percent proficiency by 2014, and it imposes consequences on schools and districts that fail to make AYP. This law is the most recent incarnation of a bipartisan standards-based reform movement that emerged from a historic 1989 summit in Charlottesville, Virginia, between President George H. W. Bush and the state governors. That meeting generated a set of national education goals that were subsequently embedded in the Clinton administration’s 1994 Goals 2000: Educate America Act. NCLB differs from that act by its far heavier emphasis on accountability and its significantly greater federal intrusion into the operations of individual schools and districts. School-based accountability programs preceded NCLB in many states. As of 2001, forty-five states published report cards on schools, and twenty-seven of them rated schools or identified low-performing ones; districts such as Chicago,

/

163

Making the Work-Based Safety Net Work Better

Dallas, and Charlotte-Mecklenburg also had full school-based accountability programs in which they rated schools on the basis of their students’ performance, provided rewards either to schools or to teachers for improved performance, and provided some combination of sanctions and assistance to poorly performing schools. Most administrative accountability systems measure only a very small fraction of the educational outcomes that stakeholders value. The outcomes in such systems typically include student achievement as measured by test scores in only the core subjects of math and reading, supplemented in some cases (as under NCLB) with some non-test-based measures such as student attendance or graduation rates. As documented by Jesse Rothstein (2007b), educational stakeholders value a much broader range of academic outcomes as well as other outcomes such as citizenship. Thus, the narrow focus of administrative accountability systems on test results in math and reading clearly privileges one narrow set of outcomes over others. Broadly speaking, two main approaches have been used to measure school performance. In “status” measures, schools are judged on the basis of their levels of performance, typically measured by average test scores or by the fraction of students attaining certain proficiency levels. In “growth” or “value-added” measures, on the other hand, schools are rated on how much they improve individual students’ performance from one year to the next (Hanushek and Raymond 2003). The simplest way to measure achievement growth, or value added, is to use gains in student test scores from one year to the next. In some cases those gains, averaged across all students in the school, are then compared to the gains in test scores that are predicted for that school given the achievement level of its students in the prior year. More refined measures of value added can be based on regression models, with or without additional statistical controls for student characteristics (Clotfelter and Ladd 1996; Ladd and Walsh 2002). States and local school districts have used variations of both the status and the growth approach (Hanushek and Raymond 2003). Although NCLB requires that schools be evaluated in terms of their “adequate yearly progress” toward the ultimate goal of 100 percent proficiency, such “progress” is measured on the basis of comparisons of aggregate levels of student performance from one year to the next. Thus, NCLB is essentially a status model rather than a growth model. The U.S Department of Education has recently given Arkansas, Delaware, Florida, North Carolina, and Tennessee the authority to introduce some growth elements into the basic approach. The two types of measures have somewhat different goals and generate somewhat different incentives, and therefore could influence disadvantaged children differentially. Status-based systems that focus on the percentage of students who achieve at proficient levels seek to encourage schools to raise performance at least to that level. This approach is appealing to many policymakers because it sets the same target for all groups of students and because it encourages schools to focus attention on the set of low-performing students who in the past may have received little attention. Status-based systems also have the advantage of being transparent. The goal of the growth model approach is to encourage schools to

164

/

School Reforms and Improved Life Outcomes

improve the performance of their students independent of the absolute level of that achievement. Such an approach is appealing to many people because of its perceived fairness. It explicitly takes into account the fact that where students end up is heavily dependent on where they start and that the starting points tend to be highly correlated on average with family background. At the same time, the use of the growth model approach may raise political concerns, both because the public may find the approach less transparent than the status approach and because some see it as a way of letting schools with low average performance off the hook. Systems using status and growth models generate different incentives in part because they lead to different rankings of schools. Many schools deemed ineffective according to their aggregate performance levels may actually have quite high “value added” and vice versa (Clotfelter and Ladd 1996; Ladd and Walsh 2002; Kane and Staiger 2002; Stiefel et al. 2005). Some accountability systems (for instance, Florida and North Carolina) encourage both high levels of performance and high test score growth, by focusing on both levels and gains (Ladd and Zelli 2002). In addition, the two approaches send different signals about which students deserve more attention. Under a status-based system designed to encourage schools to raise student performance to some threshold level, the position of the threshold matters. A challenging performance threshold—one that would be consistent with the high aspirations of the standards-based reform movement, for example— would provide incentives for schools to focus attention on a larger group of students than would be the case with a lower threshold. Evaluating schools on the basis of “value added,” by contrast, provides a stronger incentive for schools to expend effort on the entire student body. In such a system, however, schools may have an incentive to focus attention on the more-advantaged students because the test score gains of those students are likely to exceed those of the less-advantaged students (Ladd and Walsh 2002; Richards and Sheu 1992). In recent work, Derek Neal and Diane Whitmore Schanzenbach (2007) demonstrate that Chicago schools subject to a status-based accountability system apparently concentrate attention on the students at the margins, to the detriment of disadvantaged students. Under either approach, random errors in the measurement of student performance can generate inconsistent rankings of schools over time—a factor that weakens incentives for improvement. That is especially true for small schools because the smaller the number of students in the school, the larger the schoolwide average measurement error, and hence the less consistent the school’s ranking is likely to be from one year to the next. Schools deemed to be improving at one point in time are often found to be declining the next year due to measurement error (Kane and Staiger 2002). The problem of measurement error is exacerbated when schools are rated by means of the growth model, because it requires test scores for more then one year. The danger is that personnel in such schools may receive such inconsistent signals from one year to the next that they have little incentive to respond in a constructive way. There are other possible reasons why accountability might not lead to performance improvements. In any monitoring situation, those being monitored face in-

/

165

Making the Work-Based Safety Net Work Better

centives to appear as effective as possible against the metric being assessed. Thus, the concern arises that teachers might teach so narrowly to the high-stakes test that little or no generalizable learning would take place (Koretz and Barron 1998), and this has been found repeatedly in the literature. Typically, though, as long as the high-stakes tests reflect material that policymakers and society consider important, teaching to the test would still be expected to improve student learning in the tested areas, at least to some extent. In a few cases, however, reported gains may be completely bogus. Brian Jacob and Steven Levitt (2003) convincingly demonstrate, for example, that a small fraction of Chicago teachers responded to accountability pressures in that city by engaging in outright cheating in order to boost measured student test performance. While researchers such as Figlio and Cecilia Rouse (2006) and Jacob (2005) show that test score gains associated with high-stakes tests are larger than with low-stakes tests, suggesting that teaching to the test occurs, these researchers still find evidence of low-stakes test score gains with accountability. In addition, schools may engage in strategies that artificially improve test scores by changing the group of students subject to the test. The most widely studied behavior of this type is the selective assignment of students to special education programs whose participants are exempt from state accountability systems. The evidence is quite clear that schools have responded to accountability pressures in this way (see, for instance, Cullen and Reback 2006; Deere and Strayer 2001; Figlio and Getzler 2006; and Jacob 2005). Though there may be some debate about whether the greater rates of classification are undesirable in all cases, they nonetheless highlight the possibility that schools are manipulating the testing pool specifically to inflate measured school performance. Figlio’s (2007) finding that some Florida schools changed their discipline and suspension patterns around the time of the testing in ways consistent with the goal of improving test takers’ average scores reinforces this concern. Schools may engage in other types of strategic behavior that affect student performance. For example, Figlio and Joshua Winicki (2005) demonstrate that schools change their meals programs to include more simple sugars and less fat at the time of the tests in an apparent attempt to raise performance on high-stakes examinations, and Patricia Anderson and Kristin Butcher (2006) find that schools subject to accountability pressure are more apt than other schools to sell soft drinks and snacks through vending machines. Finally, Boyd et al. (2006) illustrate how high-stakes testing in certain grades in New York altered which teacher taught in particular grades and schools. It is not clear, however, whether the altered distribution of teachers should be interpreted as strategic behavior on the part of school officials designed to affect test scores. Importantly, many of these behaviors are less likely to occur in growth model accountability systems than in status-based systems. The reason is that in the growth approach, the manipulative behavior that increases student achievement in one year would make it more difficult for the school to attain accountability goals the following year. No such trade-off arises in status-based accountability systems. Given that the evidence suggests that manipulative behavior is more likely to affect disadvantaged students, this indicates that, at least along this di-

166

/

School Reforms and Improved Life Outcomes

mension, disadvantaged students may be better served under growth model accountability systems. The body of evidence concerning the estimated effects of accountability on student performance suggests several key conclusions. For one, most studies, including Figlio and Rouse (2006) and others, find stronger estimated accountability effects for math than for reading. The larger effects for math are intuitively plausible and are consistent with findings from other policy interventions such as voucher programs and tax and expenditure limitations. Compared to reading skills, math skills are more likely to be learned in the classroom, the curriculum is well-defined and sequenced, and there is less opportunity for input by parents to substitute for what goes on and is learned in the classroom (Cronin 2005). That said, researchers often find strong positive effects for reading as well. Jacob (2005), for instance, finds stronger accountability-related test score gains in reading than in math for low-performing students. And Rouse et al. (2007) indicate strong and lasting benefits of school accountability in both reading and mathematics, though mathematics scores still are more sensitive to accountability pressures. Thus, solid evidence exists to suggest that school accountability may improve student test performance, at least in the measured subjects, at low cost. That said, the likely effects of accountability tend to be more modest than the estimated effects of class size reductions found in the analysis of Project STAR. This may be because all accountability estimates are measured after only a single year of accountability pressure; it may take more time for schools to more fully react to the pressures inherent in accountability systems.

School Choice An alternative mechanism for providing market discipline for schools involves enhanced school choice. The rationale behind school choice involves the fact that historically, students have been assigned to schools on the basis of residential location. In such a situation, schools may feel less of an incentive to be responsive to the needs of the students in their charge because the mechanism through which families could discipline schools would involve physically moving—a costly, if not impossible, process for most families. Because of historical idiosyncrasies, in some areas of the United States it is easier for a family to move in order to change than in others. Some metropolitan areas, such as Boston or Pittsburgh, offer hundreds of school districts within a relatively small geographical area. Presumably, people moving to one of these cities could choose among these schooling options, and those already resident in one district would not have to move far to change school districts. In these cases, the argument goes, the threat of family mobility and choice forces school districts to be more responsive to their clients. But many other metropolitan areas offer much fewer public sector choices—and in the case of some southern metropolitan areas, there may be no choice at all within the public sector. In these situations families have no ability to “vote with their feet,” and school districts would be expected to be less responsive to their needs.

/

167

Making the Work-Based Safety Net Work Better

There exists some evidence that suggests that communities with more options for “traditional school choice” (that is, “voting with one’s feet” and physically moving to another local jurisdiction) have higher-performing public schools than do those with fewer options. Caroline Hoxby (2000) demonstrates that metropolitan areas with less-concentrated public schooling choices—that is, more choice among public school districts—tend to have higher test scores and lower school costs than do those with more concentrated public schooling choices. Even though a reanalysis of these data by Rothstein (2007b) suggests that these results may be overstated, at a minimum they indicate that increased traditional school choice has the potential to improve public school efficiency. But apart from the efficiency benefits of traditional school choice in general, it is much less likely that a greater number of school districts in a community will benefit disadvantaged students, and it may well harm them, for several reasons. First, in communities with large numbers of school districts, the overwhelming majority of these school districts serve middle- to upper-class families, and few, if any, housing opportunities for disadvantaged families exist in most of the school districts. Therefore, in a system in which people are residentially segregated along socioeconomic lines, the families that public school systems are more likely to be responsive to are the more affluent families with more residential, and hence school, choices. If disadvantaged families lack the liquidity to change residences, or the information necessary to fully appraise their (often limited) school options, then they might also be less likely than would more affluent families to take advantage of these options and would consequently be less likely to be informed about competing school districts. Moreover, the same choice mechanisms—voting with one’s feet, known in economics as “Tiebout sorting”—that could generate efficiency gains in public education lead to further social stratification. Rothstein (2006) presents evidence that families tend to select the same schools as others in their peer groups rather than according to academic factors, which could mean that more choice for the affluent leads to disadvantaged students being left together in separate schools. Miguel Urquiola (2005), too, demonstrates that increased levels of traditional school choice tend to lead to an increased degree of social sorting. Thus, disadvantaged families, faced with increased levels of traditional school choice, may be left holding the bag. In part because of the shortcomings of traditional school choice and the recognition that disadvantaged families may in reality be presented with few if any options for their children’s schooling, in recent years policymakers have experimented with alternative mechanisms of school choice, most notably school vouchers and charter schools. School voucher proposals can take many forms, but typically provide full or near-full coverage of tuition and fees for students to attend either private religious or nonreligious schools, and some programs, such as Florida’s voucher program, incorporate increased public school choice as well. Charter schools draw students from throughout a school district (and in some states, other school districts as well), tend to have more autonomy than do traditional public schools, and are free of many of the state and local regulations that public schools typically face. Most important, charter

168

/

School Reforms and Improved Life Outcomes

schools are schools of choice, so parents must actively elect to have their children attend these schools. Many economists find school vouchers and charter schools appealing because they increase schooling options for families whose choices might otherwise be constrained, whether by low incomes, job location, residential segregation, or other factors. If parents are well-informed about their schooling options, vouchers and charter schools have the potential to improve their children’s outcomes by giving the families increased ability to sort into optimal schools. At the same time, vouchers and charter schools can foster competition among schools, leading to improvements in the overall public sector. When a student uses a voucher to attend a private school or elects to attend a charter school, the local public school’s funding may be decreased, and the threat of budget cuts or personnel cuts could provide public schools with an incentive to improve. Recent school voucher and public school choice programs, including Florida’s Opportunity Scholarship Program and the public school choice program embedded within NCLB, have also been justified on the grounds that students should be provided a means of exiting low-quality schools so that they do not experience failure. Choice programs built around this premise would provide vouchers to all students attending low-performing schools regardless of whether they are economically constrained. Economists, however, are typically less comfortable with this type of program because it suggests that the state is better than parents at assessing school quality, and it is unclear why vouchers would lead to better schooling choices among parents who are not liquidity-constrained yet voluntarily select low-quality schools. An advantage of this type of program, however, is that failing schools face the prospect of losing all of their students, and not just the economically disadvantaged ones, which presumably provides a stronger incentive to improve. In practice, it turns out that schools typically considered failing are generally populated nearly exclusively by disadvantaged students, so some of the conceptual points are practically moot. It is clear that school choice and school accountability must coexist to some degree. The conceptual argument behind school choice requires that parents be well informed about choices at their disposal. Therefore, school choice absent some form of quality school accountability could lead to false choices. Mark Schneider, Paul Teske, and Melissa Marschall (2002) demonstrate that families, particularly disadvantaged ones, often use poor-quality information to make schooling choices when offered those choices; improving the information available to those families could improve the choices that families make. Some school choice programs, including Florida’s Opportunity Scholarships vouchers and the public school choice options mandated by NCLB, are directly embedded within the school accountability system. Florida, for instance, offers school vouchers to students zoned for schools that receive failing grades in two years in a four-year window. Such programs are justified in that they offer escape options to families forced to attend schools that are considered to be persistently failing. But, as Figlio and Marianne Page (2003) point out, directly linking school choice to school accountability may not help the most liquidity-constrained families. An optimal

/

169

Making the Work-Based Safety Net Work Better

school choice system would provide the most choice to disadvantaged students, whereas an optimal school accountability system would likely focus at least as much on measures of “value added” as on student test score levels. Figlio and Page argue that school choice and school accountability systems should therefore likely be decoupled. Largely because of the relative lack of experience with school vouchers in the United States, there exists little empirical evidence to date on the relative benefits of private school vouchers. Much of the available evidence on the efficacy of school vouchers comes from analyses of small-scale privately funded school voucher programs in Dayton, Ohio; Washington, D.C.; and New York City. (Other privately funded voucher programs in Cleveland, Indianapolis, and San Antonio have also been evaluated, but there exist more significant methodological problems.) These experiments, the results of which are reported by William Howell and Paul Peterson (2002) and Peterson et al. (2006), are noteworthy in that they involve the random assignment of children to voucher and nonvoucher groups. Families who applied for vouchers under these programs completed extensive questionnaires and the program administrators took pains to test all study participants, regardless of whether they received a voucher, over time. (In practice, however, the testing rates after the first year ranged from 40 to 66 percent, which could introduce biases into the results.) The weight of the evidence from these voucher experiment-based studies indicates that students who received vouchers experienced few generalized achievement gains as a consequence of the vouchers, though the authors did find evidence of improvements in civic behavior by voucher-participating families. The sole instance of large estimated achievement benefits from voucher participation occurs in the African American community, with consistently estimated gains in New York City and more spotty evidence in Washington, D.C. However, a reanalysis of the New York City data by Krueger and Zhu (2004) calls into question the strength of those findings. Regardless of which estimate of the voucher effects for African Americans is correct, it is incorrect to interpret these findings as the effect of the voucher offer per se. Instead, it is probably better thought of as an estimate of the shift from a public to a private school—and would be consistent with the general finding in the literature of modest positive benefits associated with private school enrollment (see, for instance, Evans and Schwab 1995; Neal 1997; Figlio and Stone 1999; Grogger and Neal 2000). And given that the take-up of vouchers was quite low— only a small fraction of eligible students applied for the vouchers, and among those who applied and received the vouchers, after three years only 53 percent of recipients in New York City and 29 percent in Washington were still enrolled in private schools—the voucher estimates in actuality reflect a nonrandom sample of low-income African Americans (Ladd 2002). Since the set of families that applied for a voucher that represented a small fraction of private school costs but a large fraction of low-income families’ incomes are themselves a very select group, this implies that the results of these voucher experiments are drawn from a select subsample of a selected subsample of the eligible population.

170

/

School Reforms and Improved Life Outcomes

Some additional evidence exists with regard to the performance of students in publicly funded voucher programs in Milwaukee and Cleveland. Studies of the Milwaukee Parental Choice Program present mixed evidence on the benefits of voucher participation, with little evidence of gains in reading and more positive (though inconsistent) evidence of gains in mathematics as a consequence of the voucher program (Rouse 1998; Witte 2000). The general consensus from the Cleveland Scholarship and Tutoring Program indicates that voucher recipients have not experienced significant performance gains, because five years after the start of the program voucher recipients’ test scores were comparable to those who applied for vouchers but did not receive them (Metcalf et al. 2003). Therefore, the weight of the evidence indicates that voucher receipt does not lead to significant gains in performance, but this conclusion should be tempered by awareness that the evidence to date is quite limited. More research is necessary before strong conclusions can be drawn about any (lack of) effect of receipt of a voucher by disadvantaged children. Even if vouchers themselves provide limited benefits to their recipients, voucher programs could provide positive benefits for disadvantaged students. Indeed, one possible reason why voucher experiment studies have found little effect of voucher receipt on student outcomes could be that the entire system of public schooling—and particularly, the public schools that stood to lose students as a result of the existence of voucher programs—improved as a direct consequence of the threat of losing students due to vouchers. Several recent studies have investigated the competitive effects of publicly funded voucher programs on public schools. Hoxby (2000), for instance, compares Milwaukee public schools to some comparison schools in Wisconsin and finds that, even if the estimated gains to voucher recipients are modest, there appears to have been a strong competitive effect on public schools in Milwaukee as a consequence of the voucher program. Ladd (2002) questions these findings, arguing that Hoxby’s sample suffers from severe selection problems. Jay Greene (2001) compares Florida schools that earned the lowest grade in 1999, when Florida’s voucher program was instituted, to those that earned the next-lowest grade, and attributes his findings of differential positive test score gains among the lowest-rated schools to the new threat of vouchers. Rajashri Chakrabarti (2007) makes a similar argument, and interprets a bottom grade in Florida as a “voucher threat” in her analysis of voucher system designs in Florida and Milwaukee. Ladd and Elizabeth Glennie (2001), however, point out that the gains in test scores in Florida were comparable in magnitude to those seen in the roll-out of an accountability system in North Carolina that stigmatized low-performing schools but did not include a voucher threat. They argue that it is premature to interpret the differential improvements of bottom-scoring schools as a response to voucher threats. Figlio and Rouse (2006) present careful causal evidence from Florida that demonstrates that the overwhelming majority of this differential performance gain was likely due to grading stigma rather than voucher threats, consistent with Ladd and Glennie’s assessment. That said, new work by Rouse et al. (2007) suggests that in the longer run, the threat of vouchers in Florida is associ-

/

171

Making the Work-Based Safety Net Work Better

ated with a positive effect on test scores in threatened schools that is likely independent of the grading stigma factor. Therefore, there exists some evidence to suggest that a voucher program, once it becomes more mature, has the potential to positively affect the performance of the schools that disadvantaged students typically attend. Scaling up the policy from a small-scale voucher program to a systemwide voucher program could lead to wildly variable results. General equilibrium theoretical models of education systems (Epple and Romano 1998; Nechyba 1999, 2000) generally predict that disadvantaged students in the public sector will experience worsened outcomes in the case of a universal voucher system, because of changes in financial factors and peer effects. Dennis Epple and Richard Romano (1998) find that some students who take vouchers to the private sector are also made worse off under a universal voucher program. Although it is impossible to directly test these models, Epple, Figlio, and Romano (2004) present empirical evidence from United States data that support the model’s key theoretical predictions that students sort into public and private schools on the basis of their family income and student ability level, and that within a given private school there is an inverse relationship between income and ability. While there exists no experience in the United States that can inform this discussion, international experience may prove helpful. Evidence from Chile’s universal school voucher program suggests that the voucher system led to an increase in the dispersion of educational outcomes, with relatively disadvantaged students likely made worse off (Hsieh and Urquiola 2006). Edward Fiske and Ladd (2000) present evidence that universal public school choice in New Zealand led to diminished educational opportunities and quality at the bottom of the educational distribution in that country. Although some disadvantaged students in Chile and New Zealand were made better off as a consequence of universal vouchers, the evidence indicates that universal vouchers increased, rather than reduced, inequality in those countries, results consistent with the general equilibrium theoretical predictions. A middle ground that would provide increased educational choice for disadvantaged families while reducing the likelihood of negative systemic effects for these families would involve a means-tested voucher program that is limited to disadvantaged families. Epple and Romano (2007) expand their theoretical model to describe optimal voucher design and show that under certain circumstances, means-tested targeted vouchers offer the opportunity for social optima to be reached. Such a policy has been implemented in Milwaukee, and even though the evidence regarding academic outcomes is relatively weak, the evidence on parental satisfaction is incontrovertible (Witte 2000). Dramatically increased parental satisfaction is evident in the evaluations of the privately funded voucher programs as well (Howell and Peterson 2002). But for any means-tested program to be successful, it must be coupled with strong and useful information about public and private school quality, and should not have income thresholds that are so high that the most disadvantaged potential recipients are, in effect, crowded out of the school choice system that was designed to benefit them.

172

/

School Reforms and Improved Life Outcomes

Charter Schools as an Alternative School Choice Mechanism More states have had experience with charter schools than with school vouchers; indeed, charter schools are now in operation in thirty-nine states and the District of Columbia, and more than twenty times as many students in the United States are enrolled in charter schools as are utilizing publicly funded school vouchers. Although the experience with charter schools is still relatively short, there have been important results from several states concerning their efficacy. Employing student-level longitudinal administrative data Tim Sass (2006), Robert Bifulco and Ladd (2006), and Hanushek et al. (2007) estimate the relative test score gains of charter school students from Florida, North Carolina, and Texas, respectively, to comparable students who did not attend charter schools. The results of these studies is highly consistent. The three sets of researchers generally find no positive effect on average of charter school participation on student test scores; if anything, they find somewhat negative estimated effects of charter schools. Hoxby and Sunali Murarka (2007), on the other hand, use data from charter school lotteries in New York City and find evidence of modest positive test score gains, especially in mathematics, associated with charter school participation. The two types of empirical results indicate both the promise and the pitfalls of charter schools as education reform. The small or negative average estimated effects of charter schools using state-level data from Florida, North Carolina, and Texas are due to the fact that most charter schools analyzed are relatively new. The authors that look at the phase-in effects of charter schools generally find more positive effects of charter schools as they mature—although the North Carolina study finds consistently negative effects of charter schools even with more mature charter schools. The charter schools studied by Hoxby and Murarka are not typical of charter schools in that they are more mature and, importantly, sufficiently sought-after to necessitate admissions lotteries. And Celeste Carruthers (2008), in her recent analysis of North Carolina charter schools, shows that as charter schools age their staffing quality more closely approximates that of public schools, whereas young charter schools tend to be staffed by rookie and less-qualified teachers than are similarly-aged public schools. Therefore, the two sets of results are more consistent than many observers believe. Taken together, the empirical evidence on charter school achievement indicates that many charter schools are mediocre at best, and that young, inexperienced charter schools could have ill effects on students. But charter schools that mature and that stand the market test could have significantly positive effects on student test scores. This should be unsurprising: The charter school movement that sought to free charter schools from many of the rules and regulations under which regular public schools operate encourages experimentation. Some experiments, however, are more successful than others, and some are better implemented than others. But this freedom to experiment can lead to both considerable successes and colossal failures (Carnoy et al. 2006; Schneider and Buckley 2006).

/

173

Making the Work-Based Safety Net Work Better

But these findings present a cautionary tale regarding the use of charter schools as a policy to bring about education reform. The principal clients of charter schools are disadvantaged students who come from families that are often poorly equipped to evaluate charter schools. If the vast majority of charter-attending disadvantaged students attend relatively new and untested charter schools, as is now the case in Florida and other states, it is questionable whether these students will gain academically, at least as measured in terms of test scores, by attending charter schools, and they may be harmed to some degree. (Unmeasured here, but very important, is that charter schools also tend to offer alternative pedagogical styles. Even if students do not benefit in terms of elevated measured test scores, they might still benefit in other, more difficult to measure, dimensions.) Bringing charter schools into school accountability systems could help to facilitate the information-sharing that would encourage the survival of successful charter schools and perhaps lead to the improvement of less successful charter schools. Charter school advocates, however, are of mixed minds about the desirability of including charter schools in regular school accountability systems, and states differ in the degree to which this inclusion occurs. Even as charter schools are developing, and maybe producing negative shortrun results with their own students, there is some evidence that suggests that charter schools provide positive competitive pressure for public schools. Sass (2006), in his study of Florida, finds that public schools tend to improve as more charter schools open nearby. And even in North Carolina, where the most negative evidence of charter school productivity has been found, Bifulco and Ladd (2006) show that not only do charter schools tend to siphon off nontrivial fractions of students from nearby public schools, therefore providing reason to believe that a competitive effect might be in play, but they also may provide modest competitive pressure for public schools. In some, but not all, of their specifications, Bifulco and Ladd find evidence of a positive spillover from charters to public schools nearby. Both of these studies are subject to criticisms about reverse causality, but they do show that charter schools have the potential to pressure the regular public schools that serve the generally disadvantaged populations from which they draw. In summary, although there are no silver bullets, school accountability and school choice both present the possibility for improved achievement among disadvantaged students at little additional cost to taxpayers. However, both of these types of policies also have the potential to have deleterious effects for disadvantaged students. In the case of accountability, design issues could lead schools to either promote the performance of disadvantaged students or further marginalize these students. And school choice plans could provide both pluses and minuses for disadvantaged students, depending on the nature of the choices provided and the quality of information made available to parents. With both of these education reforms, the devil is certainly in the details. But regardless, one should not expect revolutionary results from either type of school reform: Even the most optimistic estimates of the benefits of school choice and school accountability amount to perhaps one-third to one-half of the test score gaps between disadvantaged and relatively advantaged students, and the majority of the evidence suggests more modest likely effects of these policies. That said, the potential for significant reductions 174

/

School Reforms and Improved Life Outcomes

in these test score gaps at relatively low cost suggests that more experimentation with these policies—along with class size reductions and alternative teacher compensation policies—appears to be in order.

REFORMING TWENTY-FIRST-CENTURY SCHOOLS The preceding discussion makes clear that improving the academic performance of disadvantaged children is a very difficult nut to crack. That said, the recent evidence on factors potentially within the purview of education policy and practice provides some reason for optimism. A number of policies apparently improve the test performance of disadvantaged children, and could form the backbone of meaningful systemic reform. Teachers are central to any education reform aimed at increasing the odds of adult self-sufficiency among disadvantaged children. There exist massive differences in teacher quality both between and within schools. School reforms that provide incentives for high-quality teachers to move to, and remain in, schools serving disadvantaged students could prove very beneficial. These policies must be carefully considered, however, as measuring what makes a quality teacher is a tricky proposition. It is clear that most measured teacher attributes are unrelated to student outcomes, so incentives for teachers to relocate to high-poverty schools could instead be based on measurements of teacher value-added. But teacher value-added is challenging to identify and is noisily measured, so great care must be taken to ensure that teachers identified as high-quality from a statistical perspective are done so consistently. The evidence also suggests that alternative pathways to teaching, such as Teach For America, should be explored, but that these teachers might benefit from additional pedagogical, curriculum design, and classroom management training. But regardless of the process, providing strong financial incentives for the right teachers to teach in low-income schools is likely to be quite beneficial. These incentives should not be universal, however, but should rather be targeted to attracting measured high-quality (or high-potential) teachers to low-income schools. Offering incentives for schools and teachers to improve could also have strong benefits. The limited existing evidence indicates that school accountability and teacher merit pay are both associated with improved student outcomes. But since the evidence strongly indicates that teachers and schools respond to incentives, it is imperative that the proper incentives be put forward. Accountability systems that provide rewards to getting students above a given threshold are less likely to benefit disadvantaged children than are accountability systems that reward schools and teachers on the basis of test score gains from one year to the next. Since teacher quality is difficult to measure, the ideal teacher merit pay system would be embedded within a school accountability system that provided both collective benefits for good performance that accrue to all teachers in the school and individual benefits for good performance that are based on objective measures of teacher performance, such as student test score gains after controlling for student background attributes. These individual benefits should likely also include a subjective component that is based on peer or principal ratings of teacher quality. It /

175

Making the Work-Based Safety Net Work Better

would be easier to truthfully elicit principal ratings of teachers if their own salaries were highly related to the typical test score gain in their own school, overall. School districts and states should be encouraged to try linking substantial principal merit pay to more modest but still consequential individual teacher merit pay. School accountability provides opportunities for both central monitoring of school performance and increased market pressures. Better information could conceivably lead to parents’ making more informed choices and thus over time to a potentially greater competitive response to vouchers or charter schools. But again, I suspect that the most important way in which this would occur involves the recruitment and retention of high-quality teachers in disadvantaged schools. It is important to be mindful of the general equilibrium effects of education policies as well. Reducing class sizes across the board could lead to improved test scores in some circles, but in practice this would provide opportunities for wealthier schools to “poach” low-income schools for quality teachers, perhaps leaving disadvantaged schools with smaller classes but much worse teachers. School choice and school accountability systems provide incentives to improve, but especially in the case of school accountability these incentives might end up inadvertently targeted to individuals at certain points in the ability or performance distribution, and may actually make disadvantaged students worse off. Just because there are few known policies that improve the performance of disadvantaged students does not mean that nothing should be done to improve policy. Instead, I strongly encourage widespread policy experimentation with teacher (and principal) pay for performance, linked teacher and school accountability, and alternative pathways into teaching. Even better would be to try to simultaneously estimate the effects of teacher incentives and the effects of class size reductions, to determine their independent contributions to student learning gains, and to gauge the degree to which class size and teacher quality are complementary. And of course, forging a more coherent linkage between K-through-twelve education policy and children’s preschool years can only help to mitigate the differences in starting values children accrue prior to beginning school.

The author gratefully appreciates the comments of Karl Scholz, Carolyn Heinrich, Diane Whitmore Schanzenbach, and other participants in the “Pathways” conference in Madison, Wisconsin, September 6 to 7, 2007, as well as those of the anonymous referees. He alone is responsible for all errors and opinions.

REFERENCES Anderson, Patricia, and Kristin Butcher. 2006. “Reading, Writing, and Refreshments: Are School Finances Contributing to Children’s Obesity?” Journal of Human Resources 41(3): 467–94. Bacolod, Marigee. 2007. “Do Alternative Opportunities Matter? The Role of Female Labor 176

/

School Reforms and Improved Life Outcomes Markets in the Decline of Teacher Quality, 1960–1990,” Review of Economics and Statistics 89(4): 737–51. Ballou, Dale. 1996. “Do Public Schools Hire the Best Applicants?” Quarterly Journal of Economics 111(1): 97–133. ———. 2001. “Pay for Performance in Public and Private Schools.” Economics of Education Review 20(1): 51–61. Ballou, Dale, and Michael Podgursky. 1997. Teacher Pay and Teacher Quality. Kalamazoo, Mich.: W. E. Upjohn Institute. Bifulco, Robert, and Helen Ladd. 2006. “The Impacts of Charter Schools on Student Achievement: Evidence from North Carolina.” Education Finance and Policy 1(1): 50–90. Bohrnstedt, George, and Brian Stecher. 2002. What We Have Learned about Class Size Reduction in California. Sacramento: California Department of Education. Boozer, Michael, and Cecilia Rouse. 2001. “Intraschool Variation in Class Size: Patterns and Implications.” Journal of Urban Economics 50(1): 163–89. Boyd, Donald, Pamela Grossman, Hamilton Lankford, and James Wyckoff. 2006. “How Changes in Entry Requirements Alter the Teacher Workforce and Affect Student Achievement.” Education Finance and Policy 1(2): 176–216. Cahalan, Margaret, Steven Ingels, Laura Burns, Michael Planty, and Bruce Daniel. 2006. United States High School Sophomores: A Twenty-Two Year Comparison, 1980–2002. Washington, D.C.: National Center for Education Statistics. Card, David, and Alan Krueger. 1992. “Does School Quality Matter? Returns to Education and the Characteristics of Public Schools in the United States.” Journal of Political Economy 100(1): 1–40. Card, David, and Abigail Payne. 2006. “School Finance Reform, the Distribution of School Spending, and the Distribution of SAT Scores.” Journal of Public Economics 83(1): 49–82. Carneiro, Pedro, and James Heckman. 2003. “Human Capital Policy.” In Inequality in America: What Role for Human Capital Policies, edited by James Heckman and Alan Krueger. Cambridge, Mass.: MIT Press. Carnoy, Martin, Rebecca Jacobsen, Lawrence Mishel, and Richard Rothstein. 2006. “Worth the Price? Weighing the Evidence on Charter School Achievement.” Education Finance and Policy 1(1): 151–61. Carruthers, Celeste. 2008. “The Qualifications and Classroom Performance of Teachers Moving to Charter Schools.” Working paper. Gainesville: University of Florida. Chakrabarti, Rajashri. 2007. “Can Increasing Private School Participation and Monetary Loss in a Voucher Program Affect Public School Performance? Evidence from Milwaukee.” Working paper. New York: Federal Reserve Bank of New York. Clark, Damon. 2003. “Politics, Markets and Schools: Quasi-Experimental Evidence on the Impact of Autonomy and Competition from a Truly Revolutionary U.K. Reform.” Working paper. Gainesville: University of Florida. Clotfelter, Charles, and Helen Ladd. 1996. “Recognizing and Rewarding Success in Public Schools.” In Holding Schools Accountable: Performance-Based Reform in Education, edited by Helen Ladd. Washington, D.C.: Brookings Institution. Clotfelter, Charles, Helen Ladd, and Jacob Vigdor. 2006. “Would Higher Salaries Keep Teachers in High-Poverty Schools? Evidence from a Policy Intervention in North Carolina.” Working paper. Hillsborough, N.C.: Duke University. /

177

Making the Work-Based Safety Net Work Better Cronin, John. 2005. The Impact of the NCLB Act on Student Achievement and Growth: 2005 Edition. Lake Oswego, Ore.: Northwest Evaluation Association. Corcoran, Sean, William Evans, and Robert Schwab. 2004. “The Changing Quality of Teachers over the Past Four Decades.” Journal of Policy Analysis and Management 23(3): 449–70. Cullen, Julie, and Randall Reback. 2006. “Tinkering Towards Accolades: School Gaming Under a Performance Accountability System.” Advances in Applied Microeconomics 14: 1–34. Dahl, Gordon, and Lance Lochner. 2006. “The Impact of Family Income on Child Achievement.” NBER working paper 11279. Cambridge, Mass.: National Bureau of Economic Research. Darling-Hammond, Linda, Deborah Holtzman, Su Jin Gatlin, and Julian Vasquez Heilig. 2006. “Does Teacher Preparation Matter? Evidence About Teacher Certification, Teach for America, and Teacher Effectiveness.” Working paper. Palo Alto, Calif.: Stanford University. Decker, Paul, Daniel Mayer, and Steven Glazerman. 2004. “The Effects of Teach For America on Students: Findings from a National Evaluation.” Working paper. Madison: University of Wisconsin, Institute for Research on Poverty. Dee, Thomas, and Benjamin Keys. 2004. “Does Merit Pay Reward Good Teachers? Evidence from a Randomized Experiment.” Journal of Policy Analysis and Management 23(3): 471–88. Deere, Donald, and Wayne Strayer. 2001. “Putting Schools to the Test: School Accountability, Incentives and Behavior.” Working paper. College Station: Texas A&M University. Downes, Thomas. 1992. “Evaluating the Impact of School Finance Reform on the Provision of Public Education: The California Case.” National Tax Journal 45(4): 405–19. Epple, Dennis, David Figlio, and Richard Romano. 2004. “Competition Between Private and Public Schools: Testing Stratification and Pricing Predictions.” Journal of Public Economics 88(7–8): 1215–45. Epple, Dennis, and Richard Romano. 1998. “Competition Between Private and Public Schools, Vouchers, and Peer Group Effects.” American Economic Review 88(1): 33–62. ———. 2007. “Educational Vouchers and Cream Skimming.” Working paper. Gainesville: University of Florida. Evans, William, and Robert Schwab. 1995. “Finishing High School and Starting College: Do Catholic Schools Make a Difference?” Quarterly Journal of Economics 110(3): 941–74. Figlio, David. 1997. “Teacher Salaries and Teacher Quality,” Economics Letters 55(2): 267–71. ———. 2007. “Boys Named Sue: Disruptive Children and Their Peers.” Education Finance and Policy 2(4): 376–94. Figlio, David, and Lawrence Getzler. 2006. “Accountability, Ability and Disability: Gaming the System?” Advances in Applied Microeconomics 14: 35–49. Figlio, David, and Lawrence Kenny. 2007. “Individual Teacher Incentives and Student Performance.” Journal of Public Economics 91(5–6): 901–14. Figlio, David, and Maurice Lucas. 2004. “Do High Grading Standards Affect Student Performance?” Journal of Public Economics 88(9–10): 1815–34. Figlio, David and Marianne Page. 2003. “Can School Choice and School Accountability Co-

178

/

School Reforms and Improved Life Outcomes exist Successfully?” In The Economics of School Choice, edited by Caroline Hoxby. Chicago, Ill.: University of Chicago Press. Figlio, David, and Cecilia Rouse. 2006. “Do Accountability and Voucher Threats Improve Low-Performing Schools?” Journal of Public Economics 90(1–2): 239–55. Figlio, David, and Joe Stone. 1999. “Are Private Schools Really Better?” In Research in Labor Economics, Volume 18, edited by Solomon Polachek. Stamford, Conn.: JAI Press. Figlio, David, and Joshua Winicki. 2005. “Food for Thought: The Effects of School Accountability Plans on School Nutrition.” Journal of Public Economics 89(2–3): 381–94. Finn, Jeremy D., and Charles M. Achilles. 1990. “Answers and Questions About Class Size: A Statewide Experiment.” American Educational Research Journal 27(3): 557–77. Fiske, Edward, and Helen Ladd. 2000. When Schools Compete: A Cautionary Tale. Washington, D.C.: Brookings Institution. Fowler, R. Clarke. 2003. “The Massachusetts Signing Bonus Program for New Teachers: A Model of Teacher Preparation Worth Copying?” Education Policy Analysis Archives 11(13). Available at: http://epaa.asu.edu/epaa/v11n13 (accessed January 20, 2009). Glewwe, Paul, Nauman Ilias, and Michael Kremer. 2004.” Teacher incentives.” Working paper. Cambridge, Mass.: Harvard University. Greene, Jay. 2001. An Evaluation of the Florida A-Plus Accountability and School Choice Program. New York: Manhattan Institute. Grogger, Jeff, and Derek Neal. 2000. “Further Evidence on the Benefits of Catholic Secondary Schooling.” Brookings-Wharton Papers on Urban Affairs 1: 151–93. Guryan, Jonathan. 2001. “Does Money Matter? Regression-Discontinuity Estimates from Education Finance Reform in Massachusetts.” NBER working paper. Cambridge, Mass.: National Bureau of Economic Research. Hanushek, Eric. 1992. “The Trade-Off Between Child Quantity and Quality.” Journal of Political Economy 100(1): 84–117. ———. 1996. “School Resources and Student Performance.” In Does Money Matter? The Effect of School Resources on Student Achievement and Adult Success, edited by Gary Burtless. Washington, D.C.: Brookings Institution. ———. 1999. “Some Findings from an Independent Evaluation of the Tennessee STAR Experiment and from other Investigations of Class Size Effects.” Educational Evaluation and Policy Analysis 21(2): 143–63. Hanushek, Eric, John Kain, Steven Rivkin, and Gregory Branch. 2007. “Charter School Quality and Parental Decision Making with School Choice” Journal of Public Economics 91(5–6): 823–48. Hanushek, Eric, and Margaret Raymond. 2003. “Lessons about the Design of State Accountability Systems.” In NCLB: The Politics and Practice of School Accountability, edited by Paul Peterson and Martin West. Washington, D.C.: Brookings Institution Press. Harris, Douglas, and Tim Sass. 2007. “Teacher Training, Teacher Quality and Student Achievement.” Working paper. Tallahassee: Florida State University. Hatry, Harry, John Greiner, and Brenda Ashford. 1994. Issues and Case Studies in Teacher Incentive Plans. Washington, D.C.: Urban Institute Press. Howell, William, Paul Peterson, Patrick Wolf, and David Campbell. 2006. The Education Gap: Vouchers and Urban Schools, 2nd ed. Washington, D.C.: Brookings Institution Press.

/

179

Making the Work-Based Safety Net Work Better Hoxby, Caroline. 2000. “Does Competition Among Public Schools Benefit Students and Taxpayers?” American Economic Review 90(5): 1209–38. Hoxby, Caroline, and Andrew Leigh. 2004. “Pulled Away or Pushed Out? Explaining the Decline of Teacher Aptitude in the United States.” American Economic Review 94(2): 236–40. Hoxby, Caroline, and Sonali Murarka. 2007. “New York City’s Charter Schools Overall Report.” Working paper. Cambridge, Mass.: National Bureau of Economic Research, Inc. Hsieh, Chang-Tai, and Miguel Urquiola. 2006. “The Effects of Generalized School Choice on Achievement and Stratification: Evidence from Chile’s Voucher Program.” Journal of Public Economics 90(8): 1477–1503. Jacob, Brian. 2005. “Accountability, Incentives and Behavior: Evidence from School Reform in Chicago.” Journal of Public Economics 89(5): 761–96. Jacob, Brian, and Steven Levitt. 2003. “Rotten Apples: An Investigation of the Prevalence and Predictors of Teacher Cheating.” Quarterly Journal of Economics 118(3): 843–78. Jepsen, Christopher, and Steven Rivkin. 2001. “What is the Tradeoff Between Smaller Classes and Teacher Quality?” NBER working paper 9205. Cambridge, Mass.: National Bureau of Economic Research, Inc. Johnson, Susan Moore. 1986.” Incentives for Teachers: What Motivates, What Matters.” Educational Administration Quarterly 22(3): 54–79. Kane, Thomas, and Douglas Staiger. 2002. “Improving School Accountability Systems.” NBER working paper. Cambridge, Mass.: National Bureau of Economic Research, Inc. Koretz, Dan, and Sheila Barron. 1998. “The Validity of Gains on the Kentucky Instructional Results Information System (KIRIS).” Working paper. Santa Monica, Calif.: RAND Corporation. Krueger, Alan. 1999. “Experimental Estimates of Education Production Functions.” The Quarterly Journal of Economics 114(2): 497–532. Krueger, Alan, and Pei Zhu. 2004. “Another Look at the New York City School Voucher Experiment.” American Behavioral Scientist 47(5): 658–98. Krueger, Alan, and Diane Whitmore. 2001. “The Effect of Attending a Small Class in the Early Grades on College Test-Taking and Middle School Test Results: Evidence from Project STAR.” Economic Journal 111(1): 1–28. ———. 2002. “Would Smaller Classes Help Close the Black-White Achievement Gap?” In Bridging the Achievement Gap, edited by John Chubb and Thomas Loveless. Washington, D.C.: Brookings Institution. Ladd, Helen. 1999. “The Dallas School Accountability and Incentive Program: An Evaluation of its Impacts on Student Outcomes.” Economics of Education Review 18(1): 1–16. ———. 2002. “School Vouchers: A Critical View.” Journal of Economic Perspectives 16(4): 3–24. Ladd, Helen, and Elizabeth Glennie. 2001. “A Replication of Jay Greene’s Voucher Effect Study Using North Carolina Data.” In School Vouchers: Evaluating the Evidence, edited by Martin Carnoy. Washington, D.C.: Economic Policy Institute. Ladd, Helen, and Randy Walsh. 2002. “Implementing Value-Added Measures of School Effectiveness: Getting the Incentives Right.” Economics of Education Review 21(1): 1–17. Ladd, Helen, and Arnaldo Zelli. 2002. “School-Based Accountability in North Carolina: The Responses of School Principals.” Educational Administration Quarterly 38(3): 494–529.

180

/

School Reforms and Improved Life Outcomes Lankford, Hamilton, Susanna Loeb, James Wyckoff, and Donald Boyd. 2002. “Analyzing Determinants of the Matching of Public School Teachers to Jobs: Estimating Compensating Differentials in Imperfect Labor Markets.” NBER working paper 9878. Cambridge, Mass.: National Bureau of Economic Research, Inc. Lavy, Victor. 2002. “Evaluating the Effect of Teachers’ Group Performance Incentives on Pupil Achievement.” Journal of Political Economy 110(4): 1286–1317. ———. 2004. “Performance Pay and Teachers’ Effort, Productivity and Grading Ethics.” NBER Working Paper 10622. Cambridge, Mass.: National Bureau of Economic Research, Inc. Lazear, Edward. 2001. “Educational Production.” Quarterly Journal of Economics 116(3): 777–803. Loeb, Susanna, and Marianne Page. 2000. “Examining the Link Between Wages and Quality in the Teacher Workforce: The Role of Alternative Labor Market Opportunities and Non-Pecuniary Variation.” Review of Economics and Statistics 82(3): 393–408. Metcalf, Kim, Stephen West, Natalie Legan, Kelli Paul and William Boone. 2003. “Evaluation of the Cleveland Scholarship and Tutoring Program.” Working paper. Bloomington: Indiana University. Muralidharan, Karthik, and Venkatesh Sundararaman. 2007. “Teacher Performance Pay: Experimental Evidence from India.” Working paper. Cambridge, Mass.: Harvard University. Murname, Richard, and David Cohen. 1986. “Merit Pay and the Evaluation Problem: Why Most Merit Pay Plans Fail and Few Survive.” Harvard Education Review 56(1): 1–17. Murray, Sheila, William Evans, and Robert Schwab. 1998. “Education Finance Reform and the Distribution of Education Resources.” American Economic Review 88(4): 789–812. Neal, Derek. 1997. “The Effects of Catholic Secondary Schooling on Educational Achievement.” Journal of Labor Economics 15(1): 98–123. Neal, Derek, and William Johnson. 1996. “The Role of Premarket Factors in Black-White Wage Differences.” Journal of Political Economy 104(5): 869–95. Neal, Derek, and Diane Whitmore Schanzenbach. 2007. “Left Behind by Design: Proficiency Counts and Test-Based Accountability.” Review of Economics and Statistics, forthcoming. Nechyba, Thomas. 1999. “School Finance Induced Migration Patterns: The Impact of Private School Vouchers.” Journal of Public Economic Theory 1(1): 5–50. ———. 2000. “Mobility, Targeting and Private School Vouchers,” American Economic Review 90(1): 130–46. Nye, Barbara, Larry Hedges, and Spyros Konstantopoulos. 2002. “Do Low Achieving Students Benefit More from Small Classes? Evidence from the Tennessee Class Size Experiment.” Educational Evaluation and Policy Analysis 24: 201–17. Princiotta, Dan, Kristin Flanagan, and Elvira Germino-Hausken. 2006. Fifth Grade: Findings from the Fifth Grade Follow Up of the Early Childhood Longitudinal Study, Kindergarten Class of 1998–99. Washington, D.C.: National Center for Education Statistics. Reed, Deborah, Kim Rueben, and Elisa Barbour. 2006. Retention of New Teachers in California. San Francisco: Public Policy Institute of California. Richards, C., and T. Sheu. 1992. “The South Carolina School Incentive Reward Program: A Policy Analysis.” Economics of Education Review 11(1): 71–86. Rivkin, Steven, Eric Hanushek, and John Kain. 2005. “Teachers, Schools and Academic Achievement.” Econometrica 73(2): 417–58.

/

181

Making the Work-Based Safety Net Work Better Rockoff, Jonah. 2004. “The Impact of Individual Teachers on Student Achievement: Evidence from Panel Data.” American Economic Review 94(2): 247–52. Rothstein, Jesse. 2006. “Good Principals or Good Peers? Parental Valuations of School Characteristics, Tiebout Equilibrium, and the Incentive Effects of Competition among Jurisdictions.” American Economic Review 96(4): 1333–50. ———. 2007a. “Do Value-Added Models Add Value? Tracking, Fixed Effects and Causal Inference.” Working paper. Princeton, N.J.: Princeton University. ———. 2007b. “Does Competition Among Public Schools Benefit Students and Taxpayers? A Comment on Hoxby (2000).” American Economic Review 97(5): 2026–37. Rouse, Cecilia. 1998. “Private School Vouchers and Student Achievement: An Evaluation of the Milwaukee Parental Choice Program.” Quarterly Journal of Economics 113(2): 553–602. Rouse, Cecilia, and Lisa Barrow. 2006. “U.S. Elementary and Secondary Schools: Equalizing Opportunity or Replicating the Status Quo?” The Future of Children 16(2): 99–123. Rouse, Cecilia, Jane Hannaway, Dan Goldhaber, and David Figlio. 2007. “Feeling the Florida Heat? How Low-Performing Schools Respond to Voucher and Accountability Pressure.” NBER working paper. Cambridge, Mass.: National Bureau of Economic Research. Roy, Joydeep. 2004. “Impact of School Finance Reform on Resource Equalization and Academic Performance: Evidence from Michigan.” Working paper. Washington, D.C.: Economic Policy Institute. Sass, Tim. 2006. “Charter Schools and Student Achievement in Florida.” Education Finance and Policy 1(1): 91–122. Scafidi, Benjamin, David Sjoquist, and Todd Stinebrickner. 2007. “Race, Poverty, and Teacher Mobility.” Economics of Education Review 26(2): 145–59. Schneider, Mark, and Jack Buckley. 2006. “Charter Schools: Hype or Hope?” Education Finance and Policy 1(1): 123–38. Schneider, Mark, Paul Teske, and Melissa Marschall. 2002. Choosing Schools: Consumer Choice and the Quality of American Schools. Princeton, N.J.: Princeton University Press. Springer, Matthew, Keke Liu, and James Guthrie. 2005. “The Impact of Education Finance Litigation Reform on Resource Distribution: Is There Anything Special About Adequacy?” Working paper. Washington, D.C.: Economic Policy Institute. Stiefel, Leanna, Amy Ellen Schwartz, Ross Rubinstein, and Jeffrey Zabel, eds. 2005. Measuring School Performance and Efficiency: Implications for Practice and Research. American Education Finance Association 2005 Yearbook. Larchmont, N.Y.: Eye on Education. Urquiola, Miguel. 2005. “Does School Choice Lead to Sorting? Evidence from Tiebout Variation.” American Economic Review 95(4): 1310–1326. West, Jerry, Kristin Denton, and Elvira Germino-Hausken. 2000. America’s Kindergartners. Washington, D.C.: National Center for Education Statistics. Whitmore Schanzenbach, Diane. 2006. “What Have Researchers Learned from Project STAR?” Brookings Papers on Education Policy series. Washington, D.C.: Brookings Institution. Witte, John. 2000. The Market Approach to Education. Princeton, N.J.: Princeton University Press.

182

/

Chapter 6 The Impact of Incarceration on the Employment Outcomes of Former Inmates: Policy Options for Fostering Self-Sufficiency Steven Raphael

T

he United States currently incarcerates its residents at a rate that is greater than that of every other country in the world. Aggregating the state and federal prison populations as well as inmates in local jails, there were 737 inmates per 100,000 U.S. residents in 2005 (International Centre for Prison Studies 2007). This compares with a world average of 166 per 100,000 and an average among European Community member states of 135. Of the approximately 2.1 million U.S. residents incarcerated in 2005, roughly 65 percent were inmates in state and federal prisons and the remaining 35 percent were held in local jails. Current U.S. incarceration rates are unusually high relative to historical figures for the United States itself. For the fifty-year period from the 1920s through the mid-1970s, the number of state and federal prisoners per 100,000 was approximately 110. Beginning in the mid-1970s, however, state prison populations grew at an unprecedented rate, nearly quadrupling between the mid-1970s and the present. At the same time the rate of incarceration in local jails more than tripled. The incidence of these increases has not been evenly distributed across the population. In particular, these increases have been concentrated among men, among the less-educated, and among minorities. Incarceration rates have increased by particularly large amounts for less-educated African American men. These disparate impacts are best illustrated by changes in the lifetime likelihood of serving prison time by birth cohort. The Bureau of Justice Statistics estimates that the lifetime likelihood of serving time in a state or federal prison for a male child born in 1974 was 2.2 percent for whites, 13.4 percent for blacks, and 4 percent for Hispan-

/

185

Making the Work-Based Safety Net Work Better

ics. The corresponding likelihoods for a male child born in 2001 increased to 5.9 percent for whites, 32.2 percent for blacks, and 17.2 percent for Hispanics (Bonczar 2003). Moreover, estimates of the proportion of less-educated black men who have ever served time have yielded odds ratios of a prior prison spell of 2 or higher (Raphael 2005; Western 2006; Pettit and Western 2004). These trends in incarceration rates do not bode well for the socioeconomic status of minority men, their potential partners and children, and, more generally, their communities. There is growing evidence that prior incarceration and conviction adversely impacts the employment prospects of former inmates (Holzer, Raphael, and Stoll 2002; Pager 2003; Western 2002). In many instances state and federal policies compound the problems that former inmates face when seeking employment or, more generally, attempting to reenter noninstitutionalized society. In addition, recent research links the increase in incarceration rates to poor marriage outcomes (Charles and Lough 2007), further impoverishment of lowincome children (Johnson 2007), behavioral problems among the children of the incarcerated (Johnson 2007), and the spread of communicable diseases such as HIV/AIDS among disproportionately impacted communities (Johnson and Raphael, forthcoming). Any policy agenda intended to promote pathways to self-sufficiency among low-income minority communities must address these collateral consequences associated with recent incarceration trends. It is hard to imagine healthy communities where in some instances up to one-third of working-age men may be incarcerated on any given day and considerably larger fractions are either under the jurisdiction of the criminal justice system in one form or another or have previously been incarcerated. Most of these men have children and many have outstanding child support orders with which they are ill-equipped to comply. The absence of these men and the effects of incarceration on their earnings clearly diminish the resources available to poor families. Moreover, there are few programs in the current configuration of the nation’s social safety net intended to foster economic self-sufficiency among single noncustodial men. This chapter documents recent incarceration trends, discusses the evidence pertaining to the employment effects of serving time, and discusses several policy options designed to limit the adverse collateral consequences of corrections policy on poor minority communities. Regarding policy proposals, I advocate for the following: • the elimination of federal bans on the participation of certain convicted felons from participation in various public assistance programs • a rationalization of federal, state, and local government employment bans that allows for greater consideration of the particulars of individual cases • legislative guidance on how employers may and may not consider the criminal history record of an applicant • state programs that incentivize the expunging of criminal history records for former inmates that exhibit sustained desistance from criminal activity and that meet other benchmarks of responsible postrelease behavior 186

/

Employment Outcomes of Former Inmates

I also assess the likely effects on crime of reducing the U.S. incarceration rate below current levels. Policy proposals designed to alleviate the collateral employment, social, and health consequences of incarceration must confront the potential trade-offs between minimizing time served and public safety. In the policy discussion that follows, I highlight the likely impact of the proposed changes on criminal offending as well as the political feasibility of many of the proposed changes. In most instances the proposed changes are unlikely to increase criminal offending, and may even aid successful reintegration and criminal desistance. Moreover, the relative frequency with which we use incarceration as punishment has increased to such an extent that there are certainly many men who are currently sentenced to serve time that pose a minimal threat to society.

RECENT INCARCERATION TRENDS Over the three decades since about 1980 the U.S. prison incarceration rate has increased to unprecedented levels. Figure 6.1 displays the number of state and federal prison inmates per 100,000 U.S. residents. Prior to the mid-1970s, the incarceration rate was stable, hovering in a narrow band around 110 inmates per 100,000. Thereafter, however, the incarceration rate increases precipitously. Between 1975 and 2005, the prison incarceration rate more than quadrupled, from a rate of 111 to 488 per 100,000. The annual incarceration rate increased by an average of 15.7 inmates per 100,000 per year during the 1980s, 16.8 inmates per year during the 1990s, and 4.3 inmates per year during the first few year of the new century. By 2005, the point-in-time population of state and federal prisoners stood at slightly over 1.4 million inmates. State and federal prisons hold those convicted for a felony offense who are sentenced to a year or more, as well as parole violators and probation violators who may ultimately serve less than a year on any given prison admission. In addition to these inmates, there are many who are held in local and county jails while awaiting adjudication, while serving time on a sentence of less than a year, or while serving time on a prison sentence in a local jail because of overcrowding in state facilities. Figure 6.2 presents the number of inmates per 100,000 held in jails. Between 1980 and 2005, the total population in U.S. jails quadrupled, from 183,988 inmates to 747,529. Relative to the resident population, the jail population increased by slightly more than three times (an increase in inmates per 100,000 residents from approximately 81 in 1980 to 252 in 2005). Behind this steady increase in incarceration rates are large flows of inmates into and out of the nation’s prisons and jails. Given the many very short stays in local jails and the fact that even immates with longer terms are generally sentenced to less than a year, one would expect large annual flows into and out of jail. What is perhaps more surprising is the number of inmates who are released from and admitted to prison each year. There certainly are many prisoners who are serving very long sentences in the nation’s penitentiaries (inmates who are most likely to be captured by point-in-time snapshots of the prison population), and in addition /

187

FIGURE 6.1

/ Prisoners in State or Federal Prison per 100,000 U.S. Residents, 1925 to 2005

600

Prisoners per 100,000

500 400 300 200 100 0 1915

1925

1935

1945

1955

1965

1975

2005

1995

1985

Source: U.S. Bureau of the Census, Bureau of Justice Statistics.

FIGURE 6.2

/

Number of Jail Inmates per 100,000 U.S. Residents, 1980 to 2005

300

Inmates per 100,000

250 200 150 100 50

Source: U.S. Bureau of the Census, Bureau of Justice Statistics. 188

/

0 20 4 05

20

02 20

00 20

98 19

96 19

94 19

92 19

90 19

88 19

86 19

84

82

19

19

19

80

0

Employment Outcomes of Former Inmates

there are many more U.S. residents who serve relatively short spells in prison or who cycle in and out of correctional institutions while serving sequential short spells over substantial portions of their adult lives. As demonstrated by Jeremy Travis (2005), nearly all inmates are eventually released from prison, most within five years of admission. Most tellingly, annual admissions to U.S. prisons have consistently hovered around one-half the size of the prison population, and roughly half of all inmates are released in any give year. In recent decades, admissions have consistently exceeded releases, resulting in sustained increases in incarceration rates. The increasing incarceration rates displayed in figures 6.1 and 6.2 do not reflect a general increase in the likelihood of becoming incarcerated, but a concentrated increase in the incarceration risk for well-defined subsegments of the population. First, although incarceration rates have risen for both genders, the overwhelming share of these increases is accounted for by increasing rates for men (Raphael and Stoll 2009). This is not surprising considering that men consistently account for over 90 percent of the incarcerated population in current and past decades. Within the adult male population however, the increase in incarceration risk has been further concentrated among relatively young men (ages twenty-five to forty) and minority men (black men in particular). Tables 6.1, 6.2, and 6.3 demonstrate how the likelihood of incarceration has changed for adult males by race, level of educational attainment, and age. The figures in the tables are based on tabulations of the 1980 and 2000 Public Use Microdata Samples (PUMS) of the U.S. Census of Population and Housing. The decennial census enumerates both the institutionalized and the noninstitutionalized population. Within the institutionalized population one can separately identify individuals residing in nonmilitary institutions. This category includes inmates of federal and state prisons, local jail inmates, residents of inpatient mental hospitals, and residents of other non-elderly institutions. I use residence in a nonmilitary institution as the principal indicator of incarceration. In previous research (Raphael 2005), I have demonstrated that estimates of the incarcerated population based on residents in nonmilitary group quarters in the census are quite close to incarceration totals from alternative sources.1 Each table presents the proportion of the respective population that is engaged in a productive activity (either employed, in school, or in the military), the proportion that is not institutionalized but idle (not employed, not in school, not in the military), and the proportion institutionalized. All figures pertain to men eighteen to fifty-five years of age. Table 6.1 presents overall estimates for men for four mutually exclusive race-ethnicity groupings. The proportion of incarcerated increased for all groups of men between 1980 and 2000. However, the absolute increase is largest for non-Hispanic black men and Hispanic men. The 2000 census indicates that roughly 9 percent of the adult black male population was incarcerated on any given day. The comparable figures for other groups are 3 percent for Hispanics, 1.4 percent for whites, and 0.6 percent for Asians. Table 6.2 reveals that the proportion incarcerated has increased the most for the least-educated men, and that this education-incarceration relationship differs sub-

/

189

Making the Work-Based Safety Net Work Better TABLE 6.1

/ Estimates of the Proportion of Men Eighteen to Fifty-Five Engaged in a Productive Activity, Noninstitutionalized and Idle, and Institutionalized, by Race-Ethnicity

1980

2000

Change, 2000 to 1980

Non-Hispanic white Employed or in schoola Idle Institutionalized

0.899 0.093 0.008

0.878 0.109 0.014

–0.021 0.016 0.006

Non-Hispanic black Employed or in schoola Idle Institutionalized

0.758 0.206 0.037

0.673 0.239 0.089

–0.085 0.033 0.052

Non-Hispanic Asian Employed or in schoola Idle Institutionalized

0.918 0.079 0.003

0.859 0.135 0.006

–0.059 0.056 0.003

Hispanic Employed or in schoola Idle Institutionalized

0.845 0.140 0.014

0.744 0.226 0.030

–0.101 0.086 0.016

Source: Author’s compilation based on 1980 and 2000 Public Use Microdata Samples (Ruggles et al. 2008). a. Includes men in the armed forces.

stantially across racial groups. Among white men in 2000, those without a high school diploma are more than twice as likely to be institutionalized relative to those with a high school diploma, with 4.5 percent of the former and approximately 2 percent of the latter institutionalized in 2000. Moreover, white male high school dropouts experienced the largest increase in institutionalization rates between 1980 and 2000 (a 2.4 percentage point increase, compared with a 1.3 percentage point increase for white high school graduates and a 0.4 percentage point increase for those with some college education). These changes as well as the levels are small in comparison to what is observed for black men. Between 1980 and 2000, the proportion of black men with less than a high school diploma who were institutionalized on any given day increases from 0.057 to 0.206. For black male high school graduates, the proportion institutionalized increases from 0.027 to 0.087. Even among black men with some college education, the incarceration increases by over two percentage points. In fact, the changes observed among this group of black men are comparable in magnitude to the changes observed among white high school dropouts. By comparison, the changes in institutionalization rates among Asian men are small, as are the changes among Hispanic men. The relatively low institutionalization rates among Hispanic men are consistent with recent research by Kirsten

190

/

0.895 0.099 0.006 0.941 0.054 0.005 0.963 0.035 0.002

High school graduate Employed or in schoola Idle Institutionalized

Some college Employed or in schoola Idle Institutionalized

College plus Employed or in schoola Idle Institutionalized 0.947 0.051 0.002

0.911 0.079 0.009

0.835 0.146 0.019

0.698 0.257 0.045

2000

0.917 0.073 0.011

0.866 0.110 0.024

0.776 0.197 0.027

0.658 0.285 0.057

1980

0.890 0.096 0.014

0.794 0.156 0.050

0.630 0.284 0.087

0.430 0.364 0.206

2000

NonHispanic Black

0.958 0.041 0.000

0.952 0.046 0.002

0.889 0.106 0.005

0.804 0.186 0.010

1980

0.913 0.087 0.000

0.880 0.115 0.005

0.793 0.195 0.012

0.699 0.278 0.023

2000

NonHispanic Asian

Source: Author’s compilation based on 1980 and 2000 Public Use Microdata Samples (Ruggles et al. 2008). a. Includes men in the armed forces.

0.794 0.185 0.021

1980

NonHispanic White

0.943 0.053 0.004

0.927 0.065 0.007

0.864 0.124 0.011

0.793 0.188 0.020

1980

0.892 0.101 0.007

0.855 0.126 0.019

0.734 0.232 0.035

0.667 0.297 0.036

2000

Hispanic

/ Estimates of the Proportion of Men Eighteen to Fifty-Five Engaged in a Productive Activity, Noninstitutionalized and Idle, and Institutionalized, by Race-Ethnicity and Education

Less than high school Employed or in schoola Idle Institutionalized

TABLE 6.2

Making the Work-Based Safety Net Work Better

F. Butcher and Anne Morrison Piehl (2006) demonstrating the relatively low levels of incarceration among recent immigrants (levels that are particularly surprising, given the much lower levels of educational attainment). Table 6.3 parses the data further for the least-educated by age. For high school dropouts and those with a high school diploma, the table presents the distribution of each group across the three possible states by race and ethnicity and by three age groups (eighteen to twenty-five years of age, twenty-six to thirty-five, and thirty-six to forty-five). The proportion institutionalized tends to be greatest for men between twenty-six and thirty-five within each education-race group. The most startling figures are those for black men in 2000. Among black men, roughly one third of high school dropouts between twenty-six and thirty-five are incarcerated on a given day, a number comparable to the proportion of this subgroup who are employed. The comparable figure for black men with a high school diploma is approximately 23 percent. More generally, the institutionalization rate increases for all of these subgroups of less-educated young men. However, the patterns for black males are particularly severe. The patterns depicted in tables 6.1, 6.2, and 6.3 are conservative estimates of the changes in incarceration for these groups, given that I am limited to data from the 2000 census. Since the time period when the data underlying the PUMS was last collected (approximately April 1999), the prison and jail populations have continued to grow, albeit at a slower rate. Between 1999 and 2006, the point-in-time prison population increased by roughly 270,000 inmates (a 20 percent increase) and over the same period the local jail population increased by 160,000 inmates (a 26 percent increase). By contrast, the U.S. population grew by roughly 8 percent over this time period. Thus it is likely that the 2010 census will reveal even starker patterns. The figures in tables 6.1, 6.2, and 6.3 display only the proportion incarcerated on a given day. Another relevant set of figures for understanding the importance of a prior incarceration in impacting self-sufficiency is the proportion of men who have ever served time. Given the high turnover in U.S. prisons discussed earlier, the drastic increases in incarceration rates experienced over the last three decades has left in its wake an increasingly large population of former inmates. The U.S. Bureau of Justice Statistics estimates that approximately 3 percent of white male adults, 20 percent of black male adults, and 8 percent of Hispanic male adults have served prison time at some point in their lives (Bonczar 2003). In an analysis of administrative records from the California Department of Corrections, I have estimated that at the close of the 1990s, over 90 percent of black male high school dropouts and 10 to 15 percent of black male high school graduates have served prison time in the state.2 Becky Pettit and Bruce Western (2004) estimate that the proportion of all African American men born between 1965 and 1969 who had been to prison by 1999 was 20.5 percent, 30.2 percent for black men without a college degree, and 58.9 percent for black men without a high school degree. Thus, less-educated minority men are considerably more likely to be incarcerated currently than at any time in the past. Moreover, given the fluidity of prison populations, the population of noninstitutionalized former inmates has grown

192

/

1980 0.784 0.188 0.028 0.783 0.186 0.032 0.823 0.161 0.016

Eighteen to twenty-five Employed or in schoola Idle Institutionalized

Twenty-six to thirty-five Employed or in schoola Idle Institutionalized

Thirty-six to forty-five Employed or in schoola Idle Institutionalized 0.666 0.286 0.047

0.683 0.249 0.069

0.797 0.161 0.041

2000

NonHispanic White

0.726 0.240 0.034

0.634 0.281 0.085

0.604 0.314 0.081

1980

0.423 0.387 0.191

0.343 0.336 0.321

0.473 0.307 0.221

2000

NonHispanic Black

0.845 0.150 0.005

0.783 0.207 0.010

0.791 0.192 0.017

1980

0.685 0.301 0.013

0.655 0.311 0.034

0.794 0.164 0.043

2000

NonHispanic Asian

Less than High School

0.645 0.318 0.038

0.672 0.289 0.039

0.703 0.257 0.039

2000

(Table continues on page 194.)

0.824 0.165 0.011

0.807 0.170 0.023

0.760 0.212 0.028

1980

Hispanic

/ Estimates of the Proportion of Men Eighteen to Fifty-Five Engaged in a Productive Activity, Noninstitutionalized and Idle, and Institutionalized, by Race-Ethnicity, Age, and Education

Age and Status

TABLE 6.3

0.900 0.093 0.007 0.926 0.069 0.005

Twenty-six to thirty-five Employed or in schoola Idle Institutionalized

Thirty-six to forty-five Employed or in schoola Idle Institutionalized 0.845 0.137 0.018

0.845 0.131 0.024

0.843 0.136 0.021

0.827 0.156 0.017

0.780 0.184 0.036

0.742 0.229 0.029

1980

0.635 0.280 0.085

0.624 0.259 0.117

0.634 0.281 0.084

2000

NonHispanic Black

0.913 0.085 0.001

0.888 0.104 0.008

0.871 0.123 0.007

1980

0.785 0.208 0.007

0.769 0.213 0.019

0.848 0.140 0.012

2000

NonHispanic Asian

High School Graduates

Source: Author’s compilation based on 1980 and 2000 Public Use Microdata Samples (Ruggles et al. 2008). a. Includes men in the armed forces.

0.872 0.121 0.007

Eighteen to twenty-five Employed or in schoola Idle Institutionalized

2000

NonHispanic White 1980

/ (Continued)

Age and Status

TABLE 6.3

0.898 0.094 0.008

0.874 0.111 0.015

0.844 0.145 0.012

1980

0.725 0.244 0.032

0.726 0.231 0.043

0.760 0.206 0.034

2000

Hispanic

Employment Outcomes of Former Inmates

continuously and now constitutes sizable minorities, and in some instances majorities, of certain subgroups of U.S. men.

HOW SERVING TIME IMPACTS ONE’S EMPLOYMENT PROSPECTS The discussion of the patterns in tables 6.1, 6.2, and 6.3 focused primarily on the changes in incarceration rates occurring between 1980 and 2000. We saw marked increases in the proportion of men incarcerated on any given day for relatively young, less-educated minority men. Conversely, there are corresponding sizable declines in the proportions of men who are active in a productive activity (defined here as in school, employed, or in the military). For example, table 6.2 reveals declines in the proportion of black men who are active between 1980 and 2000 of twenty-three percentage points for high school dropouts, fifteen percentage points for high school graduates, and seven percentage points for those with some college education. These declines are particularly large for the young and less-educated minority men profiled in table 6.3. In previous research (Raphael 2007b), I have found that these declines were sufficient to drive black male employment rates below corresponding black female employment rates for all educational attainment groups with the exception of college graduates (where the inter-gender difference is zero). This pattern is not observed among African Americans in previous years nor among any of the other racial-ethnic groups listed in the tables. Figure 6.3 demonstrates directly the correspondence between the changes in the proportion who are employed-active and the changes in the proportion who are incarcerated. The figure plots the ten-year changes in the proportion who are employed-active for the 1980s and 1990s against the corresponding ten-year changes in the proportion who are institutionalized for each of the demographic groups defined by the complete interaction of the four race-ethnicity groups and four education-attainment groups displayed in table 6.2 as well as four age groups corresponding to those used in table 6.3 plus the group of men forty-six to fifty-five. There is a clear negative correlation between these two variables. The results from a simple bivariate regression suggest that a 1-percentage-point increase in the proportion incarcerated is associated with a 0.83-percentage-point decrease in the proportion active. If one were to interpret this coefficient as a causal effect, it would suggest that the twenty-four-percentage-point increase in the incarceration rate of male black high school dropouts between twenty-six and thirty-five caused an approximate twenty-percentage-point decline in the employment rate of this group (thus explaining almost 70 percent of the actual decline of twentynine percentage points). What causal pathways may link changes in incarceration rates to the employment outcomes of low-skilled men? First, there is a simple contemporaneous mechanical incapacitation effect of incarceration, in that institutionalized men cannot be employed in a conventional manner. If one were to randomly select a group of

/

195

Making the Work-Based Safety Net Work Better FIGURE 6.3

/ Scatter Plot of Change in the Ten-Year Changes in the Proportion Employed, in School, or in the Military Against the Ten-Year Change in the Proportion Institutionalized, 1980 to 2000

Change in Proportion Active

0.05

Change in Proportion Active = –0.031 – 0.829* Change in Proportion Incarceration, R2 = 0.207 standard errors

0 −0.1

–0.05

0.05

0.1

0.15

–0.05

–0.1

–0.15

–0.2 Change in Proportion Institutionalized Source: Author’s compilation based on 1980 and 2000 Public Use Microdata Samples (Ruggles et al. 2008).

men and incarcerate them, the slope coefficient from a regression of the change in employment on the change in incarceration should equal the employment rate for men overall. To be sure, those admitted to prison are hardly a random sample of adult men and are likely to have employment rates substantially below that of the average male. Nonetheless, exogenous increases in incarceration will mechanically reduce the employment rate for those impacted to the extent that some of the newly admitted inmates were employed at the time of arrest.3 Since the lion’s share of the increase in incarceration since the mid-1970s reflects changes in sentencing policy rather than changes in behavior,4 this short-term contemporaneous effect will be particularly important for the most impacted subgroups.5 Beyond this contemporaneous effect, incarceration is also likely to have a dynamic lagged impact on the employment prospects of former inmates as well as a contemporaneous impact on the employment outcomes of men who have not been to prison yet come from demographic subgroups with high incarceration rates. On the positive side, a spell in prison may straighten some men out, and the

196

/

Employment Outcomes of Former Inmates

desire to avoid future prison spells may lead some to focus on living a conventional law-abiding life. Such a positive impact is akin to what criminologists refer to as a specific deterrent effect of incarceration, and may ultimately increase the employability of former inmates. On the negative side, inmates fail to accumulate human capital while incarcerated and may experience an erosion of prosocial tendencies, and may perhaps experience the enhancement of antisocial attitudes and a propensity toward violence. Moreover, the stigmatizing effects (sometimes exacerbated by state and federal policy) associated with a prior felony conviction and incarceration faced by all former inmates is certainly an obstacle they face while searching for a job. There is a further avenue, other than the mechanical, by which incarceration may contemporaneously impact the employment prospects of low-skilled minority men. Employers may statistically discriminate against men from high incarceration demographic groups in an attempt to avoid hiring ex-offenders. All of these pathways are likely to suppress the current and future employment and earnings of men from demographic groups with high incarceration rates. This impact adversely affects the material well-being of the men directly impacted as well as of those intimates and children whose welfare is determined interdependently.

Incarceration and the Accumulation of Work Experience Serving time interrupts one’s work career. The extent of this interruption depends on both the expected amount of time served on a typical term as well as the likelihood of serving subsequent prison terms. The average prisoner admitted during the late 1990s on a new commitment faced a maximum sentence of three years and a minimum of one year, with many serving time closer to the minimum (Raphael and Stoll 2005). If this were the only time served for most, then the time interruption of prison would not be that substantial.6 However, many people serve multiple terms in prison, either due to the commission of new felonies or due to violation of parole conditions post-release. A large body of criminological research consistently finds that nearly two-thirds of ex-inmates are rearrested within a few years of release from prison (Petersilia 2003). Moreover, a sizable majority of the re-arrested will serve subsequent prison terms. Thus, for many offenders the typical experience between the ages of eighteen and thirty is characterized by multiple short prison spells with intermittent, and relatively short, spells outside of prison. In prior longitudinal research on young offenders entering the California state prison system, I documented the degree to which prison interrupts the early potential work careers of young men. I followed a cohort of young men entering the state prison system in 1990 and gauged the amount of time served over the subsequent decade (Raphael 2005). This analysis found that the median inmate served 2.8 years during the 1990s, with the median white inmate (3.09 years) and median black inmate (3.53 years) serving more time and the median Hispanic inmate (2.23

/

197

Making the Work-Based Safety Net Work Better

years) serving less time.7 Roughly 25 percent served at least 5 years during the 1990s and another 25 percent served less than 1.5 years. However, these figures are misleading as a gauge of the extent of the temporal interruption. Cumulative time served does not account for the short periods of time between prison spells where inmates may find employment, yet are not able to solidify the employment match with any measurable amount of job tenure. A more appropriate measure of the degree to which incarceration impedes experience accumulation would be the time between the date of admission to prison for the first term served and the date of release from the last term. Using time lapsed between first admission and final release during the 1990s, I found that five years elapses between the first date of admission and the last date of release for the median inmate. For median white, black, and Hispanic inmates, the comparable figures are 6.2, 6.5, and 3.2 years, respectively. For approximately one-quarter of inmates, nine years pass between their initial commission to prison and their last release. In other words, one quarter of these inmates spend almost the entire decade cycling in and out of prison. These figures for California are comparable to the characteristics of parolees receiving employment services from the New York–based Center for Employment Opportunities (CEO) whose transitional employment program is currently being evaluated by the MDRC (Bloom et al. 2007). Among participant parolees in the recent experimental evaluation of CEO, Dan Bloom et al. (2007) document an average age of 33.7, and an average lifetime cumulative time served in state prison of nearly five years. Spending five or more years of one’s early life cycling in and out of institutions must impact one’s earnings prospects. Clearly, being behind bars and spending short spans of time outside of prison prohibits the accumulation of job experiences during a period of one’s life when the returns to experience are the greatest.

Does Having Been in Prison Stigmatize Ex-Offenders? The potential impact of serving time on future labor-market prospects extends beyond the failure to accumulate work experience. Employers are averse to hiring former prison inmates and often use formal and informal screening tools to weed ex-offenders out of the applicant pool. Given the high proportion of low-skilled men with prison time on their criminal history records, such employer sentiments and screening practices represent an increasingly important employment barrier, especially for low-skilled African American men. Employers consider criminal history records when screening job applicants for a number of reasons. For starters, certain occupations are closed to felons under local, state, and in some instances federal law (Hahn 1991). In many states employers can be held liable for the criminal actions of their employees. Under the theory of negligent hiring, employers can be required to pay punitive damages as well as damages for loss, pain, and suffering for acts committed by an employee on the job (Craig 1987). Finally, employers looking to fill jobs where employee monitoring is imperfect may place a premium on trustworthiness and screen applicants accordingly.

198

/

Employment Outcomes of Former Inmates

In all known employer surveys where employers are asked about their willingness to hire ex-offenders, employer responses reveal a strong aversion to hiring applicants with criminal records (Holzer, Raphael, and Stoll 2006, 2007; Pager 2003). For example, over 60 percent of employers surveyed in the Multi-City Study of Urban Inequality indicated that they would “probably not” or “definitely not” hire applicants with criminal records, with “probably not” being the modal response. By contrast, only 8 percent responded similarly when queried about their willingness to hire current and former welfare recipients. The ability of employers to act on an aversion to ex-offenders, and the nature of the action in terms of hiring and screening behavior, will depend on employer accessibility to criminal record information. If an employer can and does access criminal records, the employer may simply screen out applicants on the basis of their actual arrest and conviction records. In the absence of a formal background check, employers may statistically profile applicants using perceived correlates of previous incarceration such as age, race, or level of educational attainment and avoid hiring those from demographic groups with high rates of involvement in the criminal justice system. Such a propensity to statistically discriminate is evident in the interaction effect of employers’ stated preference regarding their willingness to hire ex-offenders, their screening behavior on this dimension, and their propensity to hire workers from high incarceration rate groups. This relationship is illustrated in figure 6.4, which reproduces some of the key findings in Harry J. Holzer, Steven Raphael, and Michael A. Stoll (2006). The figure presents tabulations of employer survey data collected in 1993 and 1994 pertaining to the proportion of employers whose most recent hire is a black male by their self-reported willingness to hire exoffenders interacted with a self-report regarding whether the employer uses criminal history background checks in screening his or her potential employees. Among employers who indicate that they are willing to hire ex-offenders, there is no statistically discernable difference in the proportion of recent hires who are black men between those who check and those who do not check criminal backgrounds. Among employers who indicate that they are unwilling to hire exoffenders, however, checking criminal background is associated with a 5.6percentage-point increase in the likelihood that the most recent hire is a black male (statistically significant at the 5 percent level).8 Thus, among those most averse to hiring former inmates, checking backgrounds actually increases the likelihood that the firm hires black males. This pattern indicates that, in the absence of such objective screening methods, employers use more informal screening tools— such as not hiring black males—to weed out potential former inmates. Holzer, Raphael, and Stoll (2006) find similar patterns with regard to employer willingness to hire other stigmatized groups of workers, such as those with large unaccounted-for gaps in their employment histories. Regarding the direct effect of stigma on former inmates themselves, the audit study by Devah Pager (2003) offers perhaps the clearest evidence of employer aversion to ex-offenders and the stigma associated with having served time in prison. The study uses male auditors matched on observable characteristics, in-

/

199

Making the Work-Based Safety Net Work Better

Proportion Whose Most Recent Hire Was a Black Male

FIGURE 6.4

/ The Proportion of Employers Whose Most Recent Hire Was a Black Male by Their Self-Stated Willingness to Hire Ex-Offenders and by Their Practice of Checking Criminal Backgrounds When Screening Applicants Checks Criminal Backgrounds Does Not Check Criminal Backgrounds

0.14 0.12

0.118

0.116 0.107

0.1 0.08 0.061 0.06 0.04 0.02 0 Willing to Hire Ex-Offenders

Not Willing to Hire Ex-Offenders

Source: Author’s compilation based on Holzer, Raphael, and Stoll 2006.

cluding age, education, general appearance, demeanor, and race, to assess the effects of prior prison experience on the likelihood that each auditor is called back for an interview. The author finds consistently sizable negative effects of prior prison experience on the likelihood of being called back by the employer, with callback rates for the auditor with prior prison time one-half that of the matched coauditor.9

Existing Research on the Employment Consequences of Incarceration In conjunction, the effects of stigma combined with the impact of incarceration on human capital accumulation, and perhaps depreciation, suggest that serving time is likely to adversely impact one’s employment prospects. Moreover, there may be a negative spill-over affect for men from high incarceration subgroups to the extent that employers wish to screen out ex-offenders and do so using informal perceived signals of criminality such as race or gaps in one’s employment history.

200

/

Employment Outcomes of Former Inmates

A growing body of empirical research investigates the effects of being convicted and serving time on postrelease employment and earnings. In nearly all of these studies, researchers analyze the pre-to-post-incarceration path of earnings and employment of those who serve time. To be sure, the principal empirical challenge in this research is to define the counterfactual path of earnings and employment for those who go to prison. Defining such a counterfactual path is difficult, considering that men tend to go prison during a time in their lives (early to midtwenties) when labor-force attachment and earnings are changing rapidly, and those who serve time are quite different from those who do not, on both observable and unobservable dimensions. The challenges of this line of research are illustrated in figures 6.5 and 6.6. To construct these figures, I identified all young men in the 1979 National Longitudinal Survey of Youth (NLSY79) who were interviewed while incarcerated (the principal gauge of serving time in these data) for the first time at the age of twenty-three or later. I then matched each of these youth to one non-incarcerated male in the sample, defined as youth who never do time during the period covered by the NLSY79. In choosing matches, I identified all never-incarcerated youth who match each incarcerated youth exactly on age, region of residence in the country, education at twenty-two years of age, and race. From these exact matches, I then chose either the match with the closest Armed Forced Qualifying Test (AFQT) score when the AFQT was available for the incarcerated youth, or a random match (among those who exact matched on observable dimensions) for incarcerated youth with no AFQT score. Each figure presents the mean of an outcome for the group of incarcerated youth or the never-incarcerated youth for years relative to the year of first incarceration (t = 0). The figure compares outcomes for the five years preceding incarceration as well as the subsequent eight-year period. Figure 6.5 compares annual weeks worked. During the pre-incarceration period, average weeks worked among future inmates and the never-incarcerated are both increasing (by 5.5 weeks among future inmates and by 8 weeks among the comparison youth). At the point of first incarceration, however, the two series diverge sharply. Among the never-incarcerated, average weeks worked continues to increase from approximately 33 weeks at year zero to 40 weeks at year five (followed by a decline in employment corresponding to the early nineties recession). Among the incarcerated there is a sharp drop in weeks worked in the first survey year following the year of first observed incarceration (to 11 weeks). The pre-incarceration peak of 22 weeks is recovered five years post-incarceration but does not rise above the pre-incarceration level during the latter eight-year period. The divergence of the incarcerated and comparison groups is illustrated by the difference in mean weeks worked during the pre-incarceration period and the post-incarceration period. For the five pre-incarceration years, the neverincarcerated work roughly 9.5 more weeks per year than the group of future inmates. In the eight post-incarceration years, this average difference increases to 17.4 weeks. Figure 6.6 shows similar patterns for average annual earnings. During the preincarceration period, the ratio of annual earnings for the comparison sample to

/

201

Making the Work-Based Safety Net Work Better FIGURE 6.5

/ Average Annual Weeks Worked for Men Who Experience Incarceration and a Matched Comparison Group Relative to Year of First Incarceration 45 40 35 Experience Incarceration Comparison Group

Weeks Worked

30 25 20 15 Average Difference = 9.5 weeks

Average Difference = 17.4 weeks

10 5 0

–6

–4

–2

0

2

4

6

8

Year Relative to First Incarceration Source: Author’s compilation based on National Longitudinal Survey of Youth 1979 (Ohio State University 2003).

the incarcerated sample is roughly 1.5. During the post-incarceration period, this ratio increases to an average of 2.6. These two figures both illustrate the difficulties faced by research on this topic. As is evident from the employment and earnings path of the treatment group, incarceration occurs at a point in the age-earnings profile of young men where labor-force attachment is strengthening and annual earnings are increasing. Simple before-after comparisons of earnings and employment among those who experience incarceration will underestimate the true consequences of having served time to the extent that earnings and employment would have increased through this period in the absence of an incarceration spell. The figure also reveals the large base disparities between those who eventually serve time and those who don’t even after having matched on a number of demographic and human capital dimensions. The comparison sample works nine more weeks and earns 50 percent more than the sample of future inmates even before the first incarcerated spell. Thus, although pre-incarceration employment and earnings dynamics are similar, this large pre-treatment disparity in average outcomes raises questions about whether the post-incarceration employment and earnings paths of non-inmates provide accurate counterfactuals for those who serve time.

202

/

Employment Outcomes of Former Inmates FIGURE 6.6 / Average Annual Earnings Among Men Who Experience Incarceration and a Matched Comparison Sample Relative to First Year of Incarceration

Average Annual Nominal Earnings

25000 Experience Incarceration Comparison Group

20000

15000 Average Ratio, comparison/ treatment = 1.51

Average Ratio, comparison/ treatment = 2.62

10000

5000

0 –6

–4

–2

0

2

4

6

8

Year Relative to Incarceration Source: Author’s compilation based on National Longitudinal Survey of Youth 1979 (OSU 2003).

Several researchers have employed a host of strategies to address these methodological challenges using data from the NLSY79.10 Bruce Western (2002) compares the earnings trajectories of NLSY79 youths who serve time to high-risk youths who do not and finds a sizable relative decline in the hourly wages of the formerly incarcerated. In previous research (Raphael 2007a), I compare the employment outcomes of NLSY79 youths who serve time early in their lives to those who serve time later in life. I find a significant and substantial negative effect of prior incarceration on annual weeks worked that corresponds in time with one’s first incarceration spell. Using the more recent NLSY97 data, Gary Sweeten and Robert Apel (2007) estimate the effects of a prior incarceration spell on various employment, educational, and criminal justice outcomes after matching youths who serve time to those who don’t, using a large number of observable variables. The authors find sizable effects of a previous incarceration on the probability of employment five years following release. The authors also find some evidence that a prior incarceration predicts future criminal activity and poorer educational outcomes. A number of studies have used administrative data on arrest and incarceration

/

203

Making the Work-Based Safety Net Work Better

matched to administrative earnings records. Joel Waldfogel (1994) and Jeffrey Grogger (1995) are among the first to pursue this research strategy. Waldfogel uses data on people who are convicted in federal court and compares pre- and post-conviction employment outcomes culled from federal parole records. The author finds the largest earnings penalties for those who serve time and those convicted of a “breach of trust” crime. Grogger (1995) uses California administrative data to study the distributed lagged effect of arrest, conviction, probation, being sentenced to jail, and being sentenced to prison on subsequent earnings and employment. The author finds that arrest has a short-lived negative effect on earnings, whereas serving a prison sentence has a more pronounced and longerlasting negative effect on earnings. A number of recent studies have used state and federal prison administrative records combined with ES-202 earnings records to analyze the pre and post employment and earnings patterns of prison inmates.11 Jeffrey R. Kling (2006) analyzes data for federal prisoners in California and state prisoners in Florida; Haeil Jung (2007) and Rosa Cho and Robert Lalonde (2005) analyze data for state prisoners in Illinois; Becky Pettit and Christopher Lyons (2007) analyze data for prisoners in Washington state; and William J. Sabol (2007) analyzes data for prisoners in Ohio. Although these studies differ from one another along a number of dimensions, there are several consistent findings. First, the ES-202 records measure very low employment and earnings among state prison inmates prior to incarceration (with roughly one-third showing positive quarterly earnings in any given quarter for the two years period preceding incarceration). Even though this is partially explained by the incompleteness of administrative data,12 these findings do suggest low labor-force participation rates among soon-to-be inmates. Second, nearly all of the studies find that employment increases above preincarceration levels immediately following release and then declines to or falls below pre-incarceration levels within a couple of years. The small post-release employment increase is likely driven by the fact that most released prisoners are conditionally released to parole authorities and must meet certain obligations, including employment search or even formal employment requirements (perhaps entailing a job more likely to be captured in ES-202 records), in order to remain in the community. Third, several studies (Cho and Lalonde 2005; Kling 2006; Jung 2007) find that the post-release increase in employment is larger for inmates who serve longer terms. However, Kling (2006) shows that this disparity does not survive controlling for differences in inmates characteristics and program participation between inmates serving shorter and longer terms (of particular importance is the difference in participation in work-release programs). Although these studies suggest a positive effect of conditional release on employment, they are generally unable to identify the effects of incarceration on the age-earnings and age-employment profiles of those who serve time. The reliance on quarterly Unemployment Insurance (UI) records renders these results particularly sensitive to any factors that are likely to impact the probability of working for an employer who complies with labor-market regulations. If employers who

204

/

Employment Outcomes of Former Inmates

participate in work-release programs or who have working relationships with labor-market intermediaries that place former inmates have a high degree of compliance, pre- and post-incarceration employment outcomes as measured by UI earnings records may not be comparable.13 In addition, these studies do not identify a comparison group who do not serve time with whom we could compare the average earnings and employment paths of those who do. Failing to account for the slope of the age-earning profile at the time of incarceration seriously distorts inferences regarding the ultimate impacts of incarceration. A final group of studies uses data from the U.S. census to estimate the partial correlation between the proportion of a given demographic that is incarcerated and the average employment outcomes of the nonincarcerated among the corresponding group (Raphael 2005; Raphael and Ronconi 2006). These studies show that the demographic subgroups that experience the largest increase in incarceration rates also experience the largest decreases in employment among the nonincarcerated.

COULD WE REDUCE INCARCERATION RATES WITHOUT INCREASING CRIME RATES? The enormous increase in incarceration rates just documented has left in its wake a large number of former inmates and huge flows (over 600,000 per year) of released prisoners who are in need of assistance if they hope to stay out of prison and turn their lives around. Thus, a broad-based strategy for fostering pathways toward self-sufficiency among low-skilled men must include a set of policies designed to ease former inmates’ reentry into noninstitutionalized society. However, to focus entirely on the problem of reentering former inmates obscures the dynamics of the growth in incarceration and alternative policy strategies that may prevent young men from entering the system in the first place. Each year the United States admits over 600,000 new inmates to prison, many of whom are beginning the process of cycling in and out of prison. The fiscal savings from staunching this inflow are obvious, as corrections spending per inmate averages around $35,000 per year (Donohue 2007). Moreover, sending fewer men away would generate the additional benefits from reducing the negative externalities not accounted for by explicit fiscal outlays. To be sure, one might pursue social policies that reduce the likelihood that young men most at risk for serving time in prison commit punishable offenses in the first place. In this realm, perhaps the largest returns to such social investment would come from policies that decrease the likelihood of dropping out of high school and increase average educational attainment among high-risk youth, factors that are strong (and likely causal) predictors of serving time (see, for example, the estimates in Lochner and Moretti 2004). There is much room and great need for creative thinking in this area. For example, New York City’s recent experimentation (through the Opportunity NYC program) with providing cash transfers to poor

/

205

Making the Work-Based Safety Net Work Better

families conditional on the educational outcomes of their children may prove to generate substantial social returns above and beyond the private returns to these children occurring through higher educational attainment levels, to the extent that this program impacts crime rates and prevents future incarceration spells. Beyond such efforts, policymakers may consider reducing the incarceration rate simply by employing sanctions other than incarceration in punishing less serious offenders, such as shorter parole terms. Less serious parole violations may be punished by fines, community service time, or further restrictions on privileges rather than more prison time. Of course, the social costs of such a change in direction would depend on the extent to which lowering the overall incarceration rate increases crime. Criminologists and economists have devoted considerable effort to studying and measuring the extent to which prison mechanically incapacitates the criminally active from offending as well as the extent to which the threat of prison deters potential offenders (Johnson and Raphael 2007; Nagin 1998; Levitt 1996; Raphael and Stoll 2005; Spelman 1994). This research reveals that on average, putting someone away for a year reduces criminal activity. Thus the social costs of reducing incarceration can be potentially large. However, research has also revealed that the marginal crime-abating effect of incarceration declines at a quite rapid rate as the incarceration rate increases (Johnson and Raphael 2007; Raphael and Stoll 2005; Liedka, Piehl, and Useem 2006). Moreover, there is some research that indicates that selective prison-release decisions that weigh the characteristics of the inmates may have very little to no effect on crime, especially in recent years (Raphael and Stoll 2005). Is there room to systematically reduce the incarceration rates of U.S. states without appreciably affecting—that is, increasing—crime rates? I begin by characterizing the offenders most recently admitted to prison and the ways the characteristics of such an inmate have changed over the last twenty years. I then present some results on how the prison-crime effect has changed as the incarceration rate has increased.

Changes in the Marginal Offender Raphael and Stoll (2009) demonstrate several facts about the recent increase in U.S. incarceration rates. First, they find substantial increases in the amount of time served, conditional on the offense committed. The increase in time served for like offenders accounts for 25 to 30 percent of the increase in incarceration rates since 1980. Second, they find a substantial increase in the likelihood of being sent to prison for committing a crime. This accounts for another 55 percent of the increase in incarceration. The study also finds that an increasing propensity to offend explains at most 15 percent of the increase in incarceration rates. In other words, the 400 percent increase in incarceration rates has been driven primarily by changes in sentencing policy (85 percent of the increase) rather than changes in criminal behavior (at most, 15 percent of the increase).

206

/

Employment Outcomes of Former Inmates FIGURE 6.7

/ Percentage of Prison Admissions Due to the Return to Custody of Parole Violators Without a New Term

50

Percentage of Admissions

45 40 35 30 25 20 15 10 5 0 1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

Source: Author’s compilation based on the National Corrections Reporting Program (U.S. Department of Justice, various years).

The relatively small contribution of changes in behavior as well as the huge policy expansion in the use of incarceration as punishment have resulted in the incarceration of increasingly less dangerous offenders. This is reflected in both changes in the characteristics of the marginal offender as well as changes in the relationship between crime and incarceration. Here I use data from the admissions files of the National Corrections Reporting Program (NCRP) to document changes in the marginal prisoner admission. I only use data for the subset of states that consistently report admissions information to the NCRP for all years between 1984 and 2002. Fortunately, the thirty states that consistently report to the NCRP account for an average of 70 percent of annual prison admissions over this time period. Thus, I am able to characterize the overwhelming majority of prison admissions. Figure 6.7 displays the percentage of prison admissions in each year that are not attributable to a new felony commitment. Most of these admissions are of individuals returned to custody for parole violations. This category of admissions has become an increasingly important source of prison admissions over the past two decades, increasing from approximately 29 percent of admissions in 1984 to over 40 percent in 2002. Figures 6.8 and 6.9 present the composition of prison admissions for new felonies and for parole violators by most serious offense. For those admitted on a

/

207

Making the Work-Based Safety Net Work Better FIGURE 6.8

/ Percentage of Prison Admissions by Main Offense for New Commitment, 1984 to 2002

Percentage of Prison Admissions

45.00

Violent

Property

Drug

Other

40.00 35.00 30.00 25.00 20.00 15.00 10.00 5.00 0.00 1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

Source: Author’s compilation based on the National Corrections Reporting Program (U.S. Department of Justice, various years).

new felony conviction, the proportion admitted for a violent or property crime declines considerably during the 1980s and then stabilizes during the 1990s. In 1984 roughly 75 percent of prison admissions are accounted for by offenders convicted of violent or property felony offenses; by 2002 this figure falls to below 60 percent. The proportional importance of drug offenders, on the other hand, increases considerably, from slightly over 10 percent in 1984 to over 30 percent by 1990, and fluctuates around 33 percent of admissions thereafter. Similar changes are observed in the original offense composition of those returned to custody without a new felony. In 1984, almost 80 percent of former inmates returned to custody were originally committed to prison for either a violent or property crime (those convicted of property crimes predominated). Drug offenders accounted for only 5 percent of this inflow. By 2002 the importance of property and violent offenders diminishes and the importance of drug offenders increases. By 2002, those originally convicted of drug offenses constitute approximately one-third of inmates returned to custody for a parole violation. I cannot infer from the NCRP data how the criminal propensities of those ad-

208

/

Employment Outcomes of Former Inmates FIGURE 6.9

/ Percentage of Prison Admissions by Original Commitment Offense for Parole Violators, 1984 to 2002

60.00

Violent

Property

Drug

Other

Percentage of Admissions

50.00 40.00 30.00 20.00 10.00 0.00 1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

Source: Author’s compilation based on the National Corrections Reporting Program (U.S. Department of Justice, various years).

mitted to prison have changed over time, but there are a few dimensions of potential criminality and offense severity that permit an assessment of how newly admitted inmates have changed on the margin. For example, the research on lifecourse involvement in criminal activity provides strong evidence that the criminally active desist from illegal activity as they age, and that the late teens and early twenties are the most criminally active period.14 Since the NCRP data include nearly complete information on the age of prison admits, we can explore whether, along this dimension, the United States has been admitting persons to prison less criminally active than in the past. Figure 6.10 presents the 25th percentile, the median, and the 75th percentile of the age distribution of those admitted to prison for each year between 1984 and 2002, and shows that there have been striking increases in the age of prison admits throughout the age distribution. The age of the admit at the 25th percentile has increased by 2.2 years over this period, the median age has increase by 5.2 years, and the age at the 75th percentile has increase by over 6 years. Thus, to the extent that older inmates are less criminally active, the United States has been admitting less dangerous offenders in recent years. We can also use the sentences received by offense category to characterize the

/

209

Making the Work-Based Safety Net Work Better FIGURE 6.10

/ Key Percentiles of the Age Distribution for Prisoner Admissions, 1984 to 2002

43

38

Age

33

28

23

18

25th percentile 50th percentile 75th percentile 1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

Source: Author’s compilation based on the National Corrections Reporting Program (U.S. Department of Justice, various years).

severity of the offense and then use changes in this variable over time to characterize the change in prison admissions over time. But such an analysis would have to account for the fact that sentencing may have changed over time. To perform such an analysis, I do the following. First, for each of seventy offense categories reported in the NCRP I tabulate the median maximum sentence handed down to prisoners admitted in 1984. Next, I assign this median value to each admission in 1984 and each admission in 2002. Assigning the typical maximum sentence in 1984 to those admitted in 2002 allows me to characterize the sentences these latter prisoners would have received under the earlier sentencing regime. Finally, I calculate the percentiles of this constructed distribution for each year for comparison. If offenders in 2002 are admitted for less serious offenses (as judged by the courts), then the distribution of sentences should have shifted toward shorter prison spells. Figure 6.11 presents the results of this exercise. The figure shows no change at the 10th percentile and an increase in sentence length at the 25th percentile (largely driven by the increased importance of drug offenders). For sentences at or above the median, however, offense severity (as measured by the maximum sentence that would have been handed down in 1984) diminishes by substantial

210

/

Employment Outcomes of Former Inmates FIGURE 6.11

/ Distribution of Maximum Sentences, Assigning the Median Sentence for Each of Seventy Offense Categories Handed Down in 1984 to Each Admission in 1984 and Each Admission in 2002

80

Maximum Sentence, in Months

72

1984 2002

70

60

60 50

30

48

48

40

36 24

24

60

36

24

20 10 0 10th

25th

50th

75th

90th

Percentile of Sentencing Distribution Source: Author’s compilation based on the National Corrections Reporting Program (U.S. Department of Justice, various years).

amounts. In conjunction with the patterns in the age distribution, this suggests that indeed the average admit in 2002 was less criminally predisposed and had committed a less serious offense than the typical admit in 1984.

Change in the Effect of Incarceration on Crime at the Margin The characteristics of the marginal prison admissions have shifted decisively toward less serious offenders. We are currently admitting increasingly older offenders for relatively less serious offenses than in years past. To what extent has this shift impacted the crime-abating effects of incarceration? Criminologists posit that changes in incarceration impact crime through two avenues: incapacitating the criminally active and deterring the potentially criminally active. Estimation of these effects has proceeded in roughly two methodological veins. A large body of criminological research has attempted to estimate the incapacitation effect directly by using inmate surveys pertaining to previous offending

/

211

Making the Work-Based Safety Net Work Better

prior to arrest to approximate the amount of the offending the inmate would have engaged in had he not been incarcerated (reviewed in Spelman 1994, 2000). A second, smaller yet growing, body of literature seeks to estimate the total effect of incarceration on crime (incapacitation plus deterrence) by studying the aggregate empirical relationship between changes in the incarceration rate and changes in the rates of reported crimes (Johnson and Raphael 2007; Liedka, Piehl, and Useem 2006; Levitt 1996). A comparison of the findings from these two bodies of research suggests that the overwhelming effect of incarceration on crime is attributable to the incapacitation of the criminally active. With this in mind, a shift toward incarcerating older, less serious, and perhaps less criminally inclined offenders is likely to translate into smaller effects of prison on crime on the margin. Indeed, the studies that have analyzed the marginal effect of incarceration over time have found this to be the case (Johnson and Raphael 2007; Raphael and Stoll 2005; Liedka, Piehl, and Useem 2006). To illustrate this finding, table 6.4 presents results from some of my research with Rucker Johnson (Johnson and Raphael 2007). The table presents estimates of the effect of a one-unit increase in the incarceration rate (expressed as inmates per 100,000) on the number of prevented reported crimes per 100,000 and the number of prevented total crimes per 100,000, accounting for incomplete reporting to the police, for two time periods, 1978 to 1990 and 1991 to 2004. The table is based on a series of state-level panel data regressions of the change in crime rates on the change in incarceration rates that uses an instrumental variables strategy to identify exogenous variation in incarceration that is not driven by reverse causal effects of changes in crime on change in prison population.15 The key characteristic distinguishing these two periods that should be kept in mind is the difference in the average incarceration rate. Between 1978 and 1990, the population-weighted average state-level incarceration rate was approximately 186. The comparable average incarceration rate between 1991 and 2004 is 396. Table 6.4 reveals enormous declines in the amount of crime averted for the average prison year served across these two time periods. A one-unit increase in the number of people incarcerated per 100,000 U.S. residents prevented approximately 30 crimes per 100,000 during the earlier period. The comparable figure for the latter period is 8.3 crimes. Note further that the composition of the crimes prevented shifts decisively toward less serious crimes. During the 1980s, the prevention of violent crime (murder, rape, robbery, and assault) accounts for 15 percent of the roughly 30 crimes prevented per inmate; during the 1990s and early 2000s violent crime accounts for only 7 percent of the 8 crimes prevented per inmate. In fact, the lion’s share of crimes prevented through incarceration during the latter period is attributable to the effect of incarceration on larceny, or nonburglary theft without contact. In this study we also found that the elasticity of crime with respect to prison (the percentage change in crime caused by a percentage change in incarceration) declines across these two time periods (from –0.8 to –0.2 for violent crime and from –0.43 to –0.25 for property crime).16 If we make the conservative assumption that the crime-prison elasticity remained constant throughout the 1990s and early

212

/

Employment Outcomes of Former Inmates TABLE 6.4

/

Effects of Crime Abatement on Reported Felony Offenses

Effect on Crimes Reported to the Police

Effect on All Crimes Accounting for Underreporting

Violent crime Murder Rape Robbery Assault

–0.038 –0.200 –2.555 0.262

–0.038 –0.615 –4.467 0.474

Property crime Burglary Larceny Motor vehicle theft

–6.769 –2.627 –2.018

–13.484 –6.553 –2.564

–13.945

–30.247

Violent crime Murder Rape Robbery Assault

–0.006 –0.021 –0.257 –0.037

–0.006 –0.065 –0.449 –0.067

Property crime Burglary Larceny Motor vehicle theft

–0.514 –1.674 –0.505

–1.024 –6.087 –0.642

Total

–4.182

–8.340

1978 to 1990

Total

1991 to 2004

Source: Johnson and Raphael 2007. Note: Figures are estimates of the crime-abating effect of a one-unit increase in the incarceration rate on the number of crimes per 100,000 residents. These estimates come from state-level panel data regressions that model the year-to-year change in crime rates as a function of the year-toyear change in state incarceration rates. The models are estimated using variation along the dynamic adjustment path of incarceration to underlying shocks to identify as an instrument for the interyear change in incarceration rates. See Johnson and Raphael (2007) for estimation details.

2000s (an assumption that would bias the marginal crime-abatement upward), we can use the latter elasticity estimate to forecast what the marginal crime-fighting effect of a one-unit increase in the incarceration rate is for these latter years. Since with a constant crime-prison elasticity the marginal absolute effect of a one-unit increase in incarceration declines with increases in incarceration,17 the increases in the incarceration rate through the late 1990s and early 2000s suggest that the

/

213

Making the Work-Based Safety Net Work Better

crime-abating effects of prison have likely declined beyond what is depicted in table 6.4. In 2004, the national incarceration rate stood at 484 inmates per 100,000. Also, there were 465 violent crimes and 3,517 property crimes reported per 100,000 U.S. residents. My research with Rucker Johnson finds that a 1 percent increase in the incarceration rate (equivalent to an increase in the incarceration rate of 4.84 inmates per 100,000 in 2004) should yield a –0.2 percent decrease in reported violent crime and a –0.25 percent decrease in reported property crime. Given the crime rate levels in 2004 and the magnitude of a 1 percent increase in the incarceration rate, these elasticity estimates imply that the effect of a one-unit increase in the incarceration rate would be to reduce reported violent crimes by 0.2 incidents and reported property crimes by 1.5 incidents. Accounting for the underreporting of crimes to the police suggests the total effect of a one-unit increase in the incarceration rate would be to prevent 0.38 violent crime per 100,000 and 4.3 property crimes per 100,000. In other words, the increases in incarceration occurring over the past decade have served to reduce the likely marginal effect of prison and crime by half relative to the estimates presented for the latter period in table 6.4. Thus, the available evidence suggests that the effect of incarceration on crime rates has declined considerably as the incarceration rate has increased.

Summary The large expansion of the prison population along the extensive (who gets admitted) as well as the intensive (for how long) margin has corresponded with notable changes in the composition of the group of individuals admitted to prison. We are currently admitting more nonviolent drug offenders, more parole violators, and considerably older men than we admitted in years past. All of these characteristics are generally associated with lower levels of offending, suggesting that the composition of our prison admissions has shifted decisively toward less serious offenders. Consistent with these patterns, I have also shown that the effect on crime of incarcerating the average prisoner declines sharply as the incarceration rate increases. Comparisons of the crime-prison effects during the late 1970s and early 1980s to the comparable effects for the 1990s and early 2000’s reveal that in recent years we have gotten less bang for our buck in terms of crime reduction. Moreover, extending the elasticity estimates to the most recent years suggests that incarcerating the marginal prisoner is currently preventing very little crime. The lion’s share of this chapter has focused on the adverse social consequences of incarceration for prior inmates in terms of the impact on their future employment prospects and their ability to be self-sufficient and productive members of society. The essay has devoted less attention to some of the other social externalities of incarceration, such as the impact of incarceration on the spread of HIV/AIDS and other communicable diseases (Johnson and Raphael 2007), the effects on the material well-being, behavioral outcomes, and intergenerational impacts on children (Johnson 2007), and the impact on political participation (Manza

214

/

Employment Outcomes of Former Inmates

and Uggen 2006). In conjunction with the annual fiscal outlays on corrections and the deadweight loss associated with raising the needed funds, the implicit costs of these social externalities indicate that the costs of incarceration are currently quite high. The analysis of the change in the composition of prison admits as well as the direct analysis of the effects of prison on crime both indicate that the social benefits are generally quite low. Hence, it is most likely the case that we are currently beyond the socially optimal level of incarceration. Moreover, the effects on crime of reducing incarceration below current levels would likely be lower than the already-low marginal effects discussed earlier. These marginal effects should be interpreted as the likely effect on crime associated with selecting an inmate at random and releasing them to noninstitutionalized society for a year. A reduction in the use of incarceration that was targeted toward less serious offenders should result in even smaller impacts on crime than that suggested by the figures given earlier. Such considerations further exacerbate the current disparity between the actual and optimal level of incarceration. Finally, one could reanalyze the question with a focus on the opportunity cost of public resources devoted to crime abatement. Specifically, given the many alternative public investments that may have crime-preventing effects (such as investments in early childhood development, or education), is the current allocation of public funds across these investments such that the marginal impacts on crime of these investments are equal? The analysis presented in John J. Donohue and Peter Siegelman (1998) suggest that this is not the case, and that greater investment in early childhood development could yield higher reductions in crime per dollar spent in comparison to the effect of marginal corrections expenditures.

POLICY OPTIONS TO IMPROVE THE EMPLOYMENT PROSPECTS OF FORMER INMATES Spending time in prison or having a prior felony conviction in one’s history is becoming an increasingly common characteristic of low-skilled workers, especially for low-skilled minority men. Although the causes of this increased interaction with the criminal-justice system are varied, the lion’s share of this development is attributable to changes in sentencing policy that have both increased the average time that an offender spends behind bars and enlarged the scope of behavior punished by a spell of incarceration (Raphael and Stoll 2009). In a recent analysis of the declining employment rates of low-skilled minority men (Raphael 2007b), I found that only a small part of these declines can be explained by declining wages, suggesting the limits of policies designed to boost take-home earnings.18 Fully addressing the employment crises for these men requires directly addressing the barriers to employment created by their official criminal past. Facilitating the successful reentry of former inmates and prior felons into noninstitutionalized society is an extremely complex challenge that will most likely require substantial investments in training, social services, employment services, and post-release monitoring (see the discussions in Petersilia 2003 and Travis

/

215

Making the Work-Based Safety Net Work Better

2005). The sheer size of this population, with roughly 600,000 inmates released each year and nearly 5 percent of the adult male population having served time, is indicative of the enormity of the challenge. Still, there are simple steps that the state and federal government could take that would not compromise public safety, yet would eliminate some of the challenges that former inmates and felons face in procuring employment and avoiding extreme poverty post-release. I list some policy proposals, beginning with perhaps the lowest-lying fruit and progressing to more complex and expensive initiatives.

Qualifying for Public Assistance Perhaps the simplest step one could take to help ex-inmates would be to reverse the summary disqualification of former inmates and those with felony convictions from participating in federal public assistance programs and from receiving financial aid for education. Currently, those with prior drug felony convictions are prohibited from receiving federal financial education assistance. Moreover, the 1996 Personal Responsibility and Work Opportunity Reconciliation Act made drug felons ineligible for Food Stamps and cash assistance for life. The legislation gave states the option to adopt the federal ban on Food Stamps and cash assistance as is, or pass legislation to modify or eliminate the ban. States are not authorized to eliminate the ban on financial aid (Legal Action Center 2004). The only possible rationale for such collateral punishment of drug offenders is the assumption that enhanced and prolonged punishment will discourage people from engaging in drug crimes. However, the deterrence effects of incarceration itself are hotly debated among those who study the determinants of crime (see, for example, Lee and McCrary 2005 and Levitt 1996), with much research suggesting that the likely effects are quite small. With this in mind, the deterrent effects of much more removed, and perhaps less salient, punishments such as a lifetime ban on Food Stamps or becoming ineligible for Pell grants must certainly generate very little by way of crime reduction. Such bans, however, do make it more difficult for released offenders to avoid extreme poverty and to turn their lives around. Financial aid through the Pell grant program is one of the main sources of assistance for those attending community college, an important source of training and secondary education for less-skilled adults. Food Stamps very effectively provide basic assistance to meet the most fundamental needs of the poor. Banning former felons from participating in these programs is, frankly, counterproductive. States that maintain complete or partial bans on participation in public assistance should drop them, and the federal governments should reverse the ban on drug offenders’ receiving educational assistance.

Lifting Employment Bans Employment bans arising from former convictions and occupational licensing restrictions should not be applied in a blanket manner but instead should be based 216

/

Employment Outcomes of Former Inmates

on the content of one’s criminal record. When used, employment bans should be based on conviction rather than arrest records. Any bans on the employment of felons mandated by law should be based on the content of one’s previous behavior as well as on the time that has elapsed since conviction. In their analysis of the consideration by prior criminal history records, the Legal Action Center (2004) found that in nearly all states there is no standard governing the consideration of prior criminal records by employers and occupational licensing agencies. In many states employers can fire anyone who is found to have a criminal record regardless of the gravity of the offense, the time since conviction, or the relevance of the past behavior to current job responsibilities. In addition, employers are generally free, when hiring, to consider and discriminate against ex-inmates on the basis of criminal history, and most states allow employers to consider arrests not leading to conviction. Holzer, Raphael, and Stoll (2006) demonstrate that most employers of lowskilled labor check criminal records in some manner—by directly asking the applicant, paying a private firm to check, or performing a direct query of the state criminal history repository—and that the proportion of employers that check has increased considerably over the decade of the 1990s. The high propensity to check, employers’ complete discretion in considering past criminal history records, and the high proportion of men with prior convictions all indicate a need for some governing standard that recognizes not only the interests of employers but also the employment needs of former inmates and those with prior convictions. With this in mind, states should prohibit the consideration of prior arrests that did not result in a conviction in the hiring or firing of an employee. Moreover, publicly mandated employment bans of former felons for specific jobs as well as licensing bans should be based on the content of specific offenses or offender characteristics. In general, a more considerate and rational process for determining the suitability of former prisoners for employment in certain occupations is needed.

Improve Job-Search Assistance We should invest more in labor-market intermediaries who specialize in the reentry employment needs of recently released inmates. A sizable minority of employers indicate that they are indeed willing to hire offenders and actually do so, as measured by recent hiring outcomes. Government and nonprofit entities devoted to workforce development often serve an important informational role in matching clients to employers, and this greatly minimizes the search costs for both parties. For a specific group of clients who face substantial stigma problems in the search for work, such job-search assistance is likely to be particularly important. Over time such intermediaries establish long-term relationships and credibility with employers and are thus more effective in placing their clients in employment. Intermediaries should be able to identify the most work-ready from the large pool of former inmates and offer up a steady supply of reentering former in/

217

Making the Work-Based Safety Net Work Better

mates who are pre-screened and likely to be solid employees of comparable quality to an employer’s average hire. Given the scale of the flow of inmates out of prison each year (on the order of 600,000), there is a large potential role for agencies and nonprofits devoted to minimizing employment search costs, prescreening workers for employers, and aiding those who are reentering in becoming ready for conventional employment. The initial results from the MDRC experimental evaluation of the New York Center for Employment Opportunities (CEO) program are suggestive of the potential effectiveness of such intermediaries and their ability to help parolees land jobs and stay our of trouble. Bloom et al. (2007) find relatively large effects on the employment rate of parolees yet modest effects on recidivism and the likelihood of being in state prison a year after intervention. However, the latter finding is somewhat misleading, since the parolees who participated in the program have a considerably low baseline recidivism rate, due to the fact that many of the referred clients had been out of prison for some time and are already less likely to re-offend. The experimental estimates restricted to recently released inmates yielded large impacts on the likelihood of being employed, rearrested, and reincarcerated in the year following the intervention. These are particularly heartening findings and suggest that specialized and experienced intermediaries can have substantial impacts on these men’s life prospects as well as on public safety.

Set Time Limits on Criminal Records States should incentivize desistance from criminal activity by expunging certain criminal records after a fixed time period has elapsed. In a recent analysis, Megan C. Kurlycheck, Robert Brame, and Shawn D. Bushway (2006) raise the important question of whether unfettered employer access to criminal records can be justified by the legitimate concerns of employers and the public. They assess whether the rate at which young offenders desist from offending with time since the last offense makes it reasonable to limit employer access to arrest and conviction information for sufficiently distant past offenses. The authors demonstrate that for a cohort of young men in Philadelphia the likelihood of a repeat offense declines precipitously with the time since the last offense. This pattern is consistent with both a causal effect of staying clean as well as a remaining population of former offenders that becomes increasingly selected with time since the last offense (to be specific, selection toward a low propensity to offend). For policy purposes, however, the exact source of this pattern is irrelevant. In light of this pattern, the authors argue that limiting employer access to criminal history records beyond a certain time period may effectively limit the collateral consequences of prison while not necessarily exposing employers and the public to significantly higher risk. This simple proposal carries many advantages. Clearly, being able to procure and retain gainful employment is virtually a precondition for the successful reintegration of former inmates into noninstitutionalized society. The expunging of

218

/

Employment Outcomes of Former Inmates

their past offenses following a predetermined period of desistance will certainly improve the labor market as well as life prospects of former offenders. Moreover, the prospect of former inmates’ having their record wiped clean after a given period of desistence provides an incentive for them to change their behavior. Nonetheless, this proposal may have negative unintended consequences if employers care about prior criminal activity and engage in indirect and imperfect screening practices. In other words, limiting an employer’s ability to access criminal history records or to ask about prior criminal convictions may not prevent employers from using various characteristics as signals of possible prior run-ins with the law when they make hiring and promotion decisions. At a minimum, employers may be able to effectively identify ex-offenders via such signals as education, where a person comes from, or through unaccounted-for gaps in an applicant’s employment history. At worst, employers may systematically discriminate against workers from groups that they perceive to have a high propensity to offend, such as young black men. This important issue of how employers may respond to limits on access is key to designing a policy that both allows employers to take into account aspects of an individual’s history that are legitimately related to assessing potential job performance, while protecting those who, through the passage of time, have demonstrated the irrelevance of their past infractions to their future performance. There are several key choice variables that should be considered in designing an information policy that achieves this balance. First and foremost among these is the length of the time limit placed on criminal history inquiries. If the limit is set too short, employers will not have confidence in formal checks and thus will employ informal screens as a supplement, undoing much of the potential benefit to ex-offenders from suppressing such information in the first place. To the extent that the limit is too long, few offenders will benefit and there will be little added incentive to stay clean with an eye on the prospect of an expunged record. Kurlycheck, Brame, and Bushway (2006) focus on the seven-year limit set in the federal statute pertaining to the trucking industry. Clearly, more research on employer hiring practices with a focus on this specific question would greatly inform this choice. A second choice variable concerns the starting point for the time period framing the criminal history record. The authors implicitly advocate for a start date corresponding to the date of the most recent conviction, arguing that since few employers have access to incarceration information, time since incarceration is irrelevant. However, one can imagine that, with the knowledge that records are purged after seven years, employers may still downgrade the applications from young men who they suspect have served some time. Knowing that a clean criminal history record check is consistent with either never having offended or having offended and potentially served time but having had no contact with the criminal justice system for the past seven years provides considerably more information than the alternative of ignoring incarceration. A third important choice variable concerns whether there are some offenses that should never be purged. One might make the argument that someone who

/

219

Making the Work-Based Safety Net Work Better

has served time for a felony sex offense should never work with children, or that workers with prior serious violent offenses should not be placed in jobs involving security. Again, a better understanding of how employers consider such mitigating factors would provide useful information for forming a viable policy prescription. The growing numbers of noninstitutionalized felons raise important policy questions regarding reintegration and ways society can ease and facilitate the transition of former offenders into productive and stable lives. Stable employment is clearly key. We should improve the prospects of former offenders and provide a positive incentive to desist from criminal offending to the extent that we can do so without substantially harming the interests of employers.

CONCLUSION The brunt of the United States’ world-leading rate of incarceration of its citizens falls disproportionately on young minority men, in particular black men. I have argued that the experience of serving time interrupts the work careers of young men and likely compounds the employment problems that many of these men were already experiencing prior to their incarceration. Moreover, the large proportions of minority men who pass through the nation’s prisons suggest that the impact on poor minority communities more generally is likely to be quite large. I have offered several policy suggestions that would aid the reentry transition for former inmates and perhaps help prevent further criminal involvement among these men. I have also argued that policymakers should be more proactive in preventing men from entering prison in the first place. Given the likely small effects of incarceration on crime at current levels, there is room to reduce the use of incarceration and increase the use of alternative sanctions without impacting the levels of serious crime.

I thank Carolyn Heinrich, John Karl Scholz, Michael Stoll, and the participants in the Institute for Research on Poverty’s conference on “Pathways to Self-Sufficiency: Getting Ahead in the Era Beyond Welfare Reform,” for their much-valued input.

NOTES 1. To gauge the validity of using the census data in this manner, in previous research (Raphael 2005) I compare estimates of the institutionalized population from the census to estimates of the incarcerated populations from other sources by race. Although the census estimates are slightly larger than estimates of the incarcerated population from the Bureau of Justice Statistics, the disparities are quite small relative

220

/

Employment Outcomes of Former Inmates

2.

3.

4.

5.

6.

7. 8. 9.

to the overall incarcerated population. The difference likely reflects the very small remaining inpatient population in U.S. mental hospitals. The California incarceration rate is quite close to the national average in most years. California does differ, however, in the dynamics of its incarcerated population. Given the disproportionate contribution of parole violators to the prison population, the typical spell in a California prison is considerably shorter than that for the nation. The high parole failure rate leads to a much larger prison admissions rate than that for the nation. A number of studies demonstrate that roughly one-third to two-thirds of inmates are employed at the time of the arrest leading to their current incarceration (see Kling 2007; Pettit and Lyons 2007; Tyler and Kling 2007; and Sabol 2007). Alfred Blumstein and Allen J. Beck (1999) and Raphael and Stoll (2007) demonstrate that most of the increase (80 to 85 percent) in state incarceration rates is attributable to increased sentence length conditional on being sent to prison and an increased propensity to punish felons with incarceration, holding offenses constant. Raphael and Stoll (2007) simulate the counterfactual crime rate that would have occurred had the state prison incarceration rate not increased above 1980 levels. The results suggest that a higher propensity to offend can explain no more than 15 percent of the overall increase. To be sure, causality may also run in the reverse direction, that is, from declining employment prospects to criminal activity to incarceration, but the evidence on this front is rather weak. First, the decline in wages of the least-skilled men between 1980 and 2000 was heavily concentrated in the 1980s, with some low-skilled men regaining lost ground during the 1990s and beyond. However, the increase in incarceration during the 1990s was equal in magnitude to the increase occurring during the 1980s, and the incarceration rate continued to increase between 2000 and 2006. Second, evidence of a behavioral increase in criminal activity is scant, with most research suggesting that the propensity to commit crime actually declined during the 1990s even after accounting for the increase in incarceration. Of course I am not saying that a year in prison is not costly. However, a year’s absence from the labor market during the beginning of one’s worklife would have only a small effect on accumulated experience. The California inmate population is roughly evenly distributed among whites, Hispanics, and blacks and is overwhelmingly male. The 4.4-percentage-point difference-in-difference relative to firms who are willing to hire black males is statistically significant at the 10 percent level of confidence. Of course, the audit evidence is subject to the critique that the demonstration of the existence of employers who discriminate against former inmates does not necessarily imply a market-level effect of this discrimination. Former inmates can adjust their supply behavior by applying only to firms willing to hire them. To the extent that the latter set of employers is larger than the unwilling-to-hire group, the ultimate impact on employment and earnings may be negligible. However, Holzer, Raphael, and Stoll (2007) find that fairly large proportions of employers express reservations about hiring former inmates. Moreover, in labor-market models with search frictions, such unwillingness may reduce the job offer arrival rate of former inmates, resulting in

/

221

Making the Work-Based Safety Net Work Better

10. 11.

12.

13.

14.

15.

16.

222

greater unemployment, lower wages when employed (to the extent that former inmates lower the minimum wage for which they would be willing to work, and a higher proportion withdrawing from the workforce. National Longitudinal Survey; see http://www.bls.gov/nls/nlsy79.htm for more information. Employment and Wage (ES-202) data are derived from reports filed by all employers subject to unemployment compensation laws, both state and federal. Industry employment and payroll information is produced both quarterly and annually for the state, labor market areas, workforce investment areas, cities and towns, and counties. See http://lmi2.detma.org/lmi/lmi_es_a.asp for further information. Kling (2006) is the only study that compares employment as measured by quarterly earnings records to inmate self-reported employment at the time of arrest. The author reports that although only 33 percent of inmates have positive earnings in the typical pre-incarceration quarter, nearly 65 percent report being employed at the time of arrest. On the basis of analysis of CPS data for comparable men, Kling concludes that most of this disparity reflects the fact that inmates are employed in informal jobs where employers are not paying Social Security taxes or paying into the Unemployment Insurance system. Robert Kornfeld and Bloom (1999) provide a detailed comparison of earnings as measured by quarterly UI records to survey data earnings as measured in the JTPA (Job Training Partnership Act, 1982) training experiments and provide estimated program effects using the two sources of data. The authors show that earnings from the UI data are systematically lower than earnings from the survey records. Relative program effects are similar in magnitude using the two sources of information, except for young men with criminal records. The UI data yield larger program effect estimates than the survey records, suggesting that for this particular group, program participation is increasing the likelihood of working for an employer that complies with reporting and tax requirements. The fact that criminal offending declines with age is a well-known empirical pattern. Grogger (2000) demonstrates that the proportion of youth in the 1979 National Longitudinal Survey of Youth who self-report involvement in criminal activity declines with age, a fact the author attributes to increasing legal opportunity costs. In a series of papers, Robert J. Sampson and John H. Laub (1993, 1997, 2003, 2005) find little evidence for a group of life-course criminal persisters in a long panel of offenders and individuals at high risk of offending as youths. These authors find evidence that certain life events, such as getting married, having children, or being steadily employed, correspond with desisting from crime in adulthood. Specifically, we use variation along the dynamic adjustment path of prison between equilibria to identify variation in incarceration that would occur despite contemporaneous variation in crime rates. Our results for the 1978-to-1990 time period are similar in magnitude to those reported in Levitt (1996), who analyzes a similar time period yet uses the effects of court orders to relieve prisoner overcrowding occurring during the 1970s and ’80s as an instrument. For details on our instrumental variables strategy, see Johnson and Raphael (2007). Note that the figures in table 6.4 are effects of levels on levels, not elasticities. Converting to elasticities facilitates comparisons with prior research. /

Employment Outcomes of Former Inmates 17. With a constant crime-prison elasticity, ε, the relationship between incarceration and crime can be written as lnC = εlnP. Solving for the level of C yields the equation C = Pε, which will be decreasing in P so long as ε < 0. 18. To be sure, policies designed to increase take-home pay may impact the likelihood of offending in the first place. Such policy may aid in limiting further growth in the prison population, to the extent that such policy initiatives as expanding the Earned Income Tax Credit for childless adults or reforming child support policy alter the relative returns to legitimate and illegitimate activities.

REFERENCES Bloom, Dan, Cindy Redcross, Janine Zweig, and Gilda Azurdia. 2007. “Transitional Jobs for Ex-Prisoners: Early Impacts from a Random Assignment Evaluation of the Center for Employment Opportunities Prisoner Reentry Program.” MDRC Working Paper. New York: MDRC. Blumstein, Alfred, and Allen J. Beck. 1999. “Population Growth in U.S. Prisons, 1980– 1996.” Crime and Justice: A Review of Research 26: 17–62. Bonczar, Thomas P. 2003. Prevalence of Imprisonment in the U.S. Population, 1974–2001. Special report, prepared for Bureau of Justice Statistics Special, NCJ 197976. Washington, D.C.: National Criminal Justice Service. Butcher, Kirsten F., and Anne Morrison Piehl. 2006. “Why Are Immigrant Incarceration Rates So Low? Evidence on Selective Immigration, Deterrence, and Deportation.” Working paper. New Jersey: Rutgers University. Charles, Kerwin K., and Ming Ching Louh. 2007. “Male Incarceration, the Marriage Market, and Female Outcomes.” Working paper. Chicago: University of Chicago. Cho, Rosa, and Robert Lalonde. 2005. “The Impact of Incarceration in State Prison on the Employment Prospects of Women.” Harris School working paper #5-10. Chicago: University of Chicago. Craig, Scott R. 1987. “Negligent Hiring: Guilt By Association.” Personnel Administrator (October): 32–34. Donohue, John J. 2007. “Assessing the Relative Benefits of Incarceration: The Overall Change of the Previous Decades and the Benefits on the Margin.” Working paper. New Haven, Conn.: Yale University. Donohue, John J., and Peter Siegelman. 1998. “Allocating Resources Among Prisons and Social Programs in the Battle Against Crime.” Journal of Legal Studies 27(1): 1–43. Grogger, Jeffrey. 1995. “The Effect of Arrest on the Employment and Earnings of Young Men.” Quarterly Journal of Economics 110(1): 51–71. ———. 2000. “An Economic Model of Recent Trends in Violence.” In The Crime Drop in America, edited by Alfred Blumstein and Joel Wallman. Cambridge: Cambridge University Press. Hahn, Jeffrey M. 1991. “Pre-Employment Information Services: Employers Beware.” Employee Relations Law Journal 17(1): 45–69. Holzer, Harry J., Steven Raphael, and Michael A. Stoll. 2006. “Perceived Criminality, Criminal Background Checks and the Racial Hiring Practices of Employers.” Journal of Law and Economics 49(2): 451–80. /

223

Making the Work-Based Safety Net Work Better ———. 2007. “The Effect of an Applicant’s Criminal History on Employer Hiring Decisions and Screening Practices: Evidence from Los Angeles.” In Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America, edited by Shawn Bushway, Michael Stoll, and David Weiman. New York: Russell Sage Foundation. International Centre for Prison Studies. 2007. “World Prison Brief.” Available at: http:// www.prisonstudies.org (accessed on January 1, 2007). Johnson, Rucker. 2007. “Intergenerational Risks of Criminal Involvement and Incarceration.” Working paper. Berkeley: University of California. Johnson, Rucker, and Steven Raphael. Forthcoming. “The Effect of Male Incarceration Dynamics on AIDS Infection Rates Among African-American Women and Men.” Journal of Law and Economics. ———. 2007. “How Much Crime Reduction Does the Marginal Prisoner Buy?” Working paper. Berkeley: University of California. Jung, Haeil. 2007. “The Effects of First Incarceration on Male Ex-Offenders’ Employment and Earnings.” Working paper. Chicago: University of Chicago. Kornfeld, Robert, and Howard Bloom. 1999. “Measuring Program Impacts on Earnings and Employment: Do Unemployment Insurance Wage Records Agree with Survey Reports of Individuals?” Journal of Labor Economics 17(1): 168–97. Kling, Jeffrey R. 2006. “Incarceration Length, Employment, and Earnings.” American Economic Review 96(3): 863–76. Kurlycheck, Megan C, Robert Brame, and Shawn D. Bushway. 2006. “Scarlet Letters and Recidivism: Does An Old Criminal Record Predict Future Offending?” Criminology and Public Policy 5(3): 483–504. Legal Action Center. 2004. “After Prison: Roadblocks to Reentry; A Report on State Legal Barriers Facing People with Criminal Records.” New York: Legal Action Center. Available at: http://lac.org/roadblocks-to-reentry (accessed March 26, 2009). Lee, David S., and Justin McCrary. 2005. “Crime, Punishment, and Myopia,” NBER working paper #11491. Washington, D.C.: National Bureau of Economic Research. Levitt, Steven D. 1996. “The Effect of Prison Population Size on Crime Rates: Evidence from Prison Overcrowding Legislation.” Quarterly Journal of Economics 111(2): 319–51. Liedka, Raymond V., Anne Morrison Piehl, and Bert Useem. 2006. “The Crime-Control Effect of Incarceration: Does Scale Matter?” Criminology and Pubic Policy 5(2): 245–76. Lochner, Lance, and Enrico Moretti. 2004. “The Effect of Education on Criminal Activity: Evidence from Prison Inmates, Arrest, and Self Reports.” American Economic Review 94(1): 155–89. Manza, Jeff, and Christopher Uggen. 2006. Locked Out: Felon Disenfranchisement and American Democracy. New York: Oxford University Press. Nagin, Daniel S. 1998. “Criminal Deterrence Research at the Outset of the Twenty-First Century.” Crime and Justice: A Review of Research 23: 1–42. Ohio State University, Center for Human Resource Research. 2003. National Longitudinal Survey of Youth, 1979. [Computer file]. ICPSR04683-v1. Washington, D.C.: U.S. Department of Commerce, Bureau of the Census, and U.S. Department of Labor Statistics [producers], 2003. Ann Arbor, Mich.: Inter-university Consortium for Political and Social Research [distributor] 2007-09-07. doi: 10.3886/ICPSR04683. Pager, Devah. 2003. “The Mark of a Criminal Record.” American Journal of Sociology 108(5): 937–75. 224

/

Employment Outcomes of Former Inmates Petersilia, Joan. 2003. When Prisoners Come Home. Oxford: Oxford University Press. Pettit, Becky, and Christopher Lyons. 2007. ”Status and the Stigma of Incarceration: The Labor Market Effects of Incarceration by Race, Class, and Criminal Involvement.” In Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America, edited by Shawn Bushway, Michael Stoll, and David Weiman. New York: Russell Sage Foundation. ———. 2004. “Mass Imprisonment and the Life Course: Race and Class Inequality in U.S. Incarceration.” American Sociological Review 69(2): 151–69. Raphael, Steven. 2005. “The Socioeconomic Status of Black Males: The Increasing Importance of Incarceration.” In Poverty, the Distribution of Income, and Public Policy, edited by Alan Auerbach, David Card, and John Quigley. New York: Russell Sage Foundation. ———. 2007a. “Early Incarceration Spells and the Transition to Adulthood.” In The Price of Independence, edited by Sheldon Danziger and Cecilia Rouse. New York: Russell Sage Foundation. ———. 2007b. “Boosting the Earnings and Employment of Low-Skilled Workers in the United States.” In A Future of Good Jobs: America’s Challenge in the Global Economy, edited by Timothy Bartik and Susan Houseman. Kalamazoo, Mich.: W. E. Upjohn Institute. Raphael, Steven, and Lucas Ronconi. 2006. “Reconciling National and Regional Estimates of the Effects of Immigration on the U.S. Labor Market: The Confounding Effects of Native Male Incarceration Trends.” Working Paper. Berkeley: University of California. Raphael, Steven, and Michael Stoll. 2005. “The Effect of Prison Releases on Regional Crime Rates.” In The Brookings-Wharton Papers on Urban Economic Affairs, edited by William G. Gale and Janet Rothenberg Pack. Volume 5. Washington, D.C.: Brookings Institution. ———. 2008. “Why Are So Many Americans in Prison?” In Do Prisons Make Us Safer? The Benefits and Costs of the Prison Boom, edited by Steven Raphael and Michael Stoll. New York: Russell Sage Foundation. Ruggles, Steven, Matthew Sobek, Trent Alexander, Catherine A. Fitch, Ronald Goeken, Patricia Kelly Hall, Miriam King, and Chad Ronnander. 2008. Integrated Public Use Microdata Series: Version 4.0 [machine-readable database]. Minneapolis: Minnesota Population Center. Sabol, William J. 2007. “Local Labor-Market Conditions and Post-Prison Employment Experiences of Offenders released from Ohio State Prisons.” In Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America, edited by Shawn Bushway, Michael Stoll, and David Weiman. New York: Russell Sage Foundation. Sampson, Robert J., and John H. Laub. 1993. Crime in the Making: Pathways and Turning Points in Life. Cambridge, Mass.: Harvard University Press. ———. 1997. “Life Course Theory of Cumulative Disadvantage and the Stability of Delinquency.” In Developmental Theories of Crime and Delinquency, edited by Terrence Thornberry. Trenton, N.J.: Transaction Publishers. ———. 2003. “Life-Course Desisters? Trajectories of Crime Among Delinquent Boys Followed to Age 70?” Criminology 41(3): 301–39. ———. 2005. “A Life-Course View of the Development of Crime.” In Developmental Criminology and Its Discontents: Trajectory of Crime from Adulthood to Old Age, edited by Robert J. Sampson and John H. Laub. The Annals of the American Academy of Political and Social Sciences Series. Philadelphia: American Academy of Political and Social Sciences. Spelman, William. 1994. Criminal Incapacitation. New York: Plenum Press. ———. 2000. “The Limited Importance of Prison Expansion.” In The Crime Drop in Amer/

225

Making the Work-Based Safety Net Work Better ica, edited by Alfred Blumstein and Joel Wallman. Cambridge: Cambridge University Press. Sweeten, Gary, and Robert Apel. 2007. “Incarceration and the Transition to Adulthood.” Working paper. Phoenix: Arizona State University. Travis, Jeremy. 2005. But They All Come Back: Facing the Challenges of Prisoner Reentry. Washington, D.C.: The Urban Institute Press. Tyler, John H., and Jeffrey R. Kling. 2007. “Prison-Based Education and Reentry into the Mainstream Labor Market.” In Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America, edited by Shawn Bushway, Michael Stoll, and David Weiman. New York: Russell Sage Foundation. U.S. Department of Justice, Bureau of Justice Statistics. Various years. National Corrections Reporting Program, 2002: [United States] [Computer file]. Conducted by U.S. Department of Commerce, Bureau of the Census. ICPSR04345-v1. Ann Arbor, Mich.: Inter-university Consortium for Political and Social Research [producer and distributor], 2006. doi:10.3886/ICPSR04345. Waldfogel, Joel. 1994. “The Effect of Criminal Convictions on Income and the Trust ‘Reposed in the Workmen.’ ” Journal of Human Resources 29(1): 62–81. Western, Bruce. 2002. “The Impact of Incarceration on Wage Mobility and Inequality.” American Sociological Review 67(4): 526–46. Western, Bruce. 2006. Punishment and Inequality in America. New York: Russell Sage Foundation.

226

/

Chapter 7 The Growing Problem of Disconnected Single Mothers Rebecca M. Blank and Brian K. Kovak

A

key issue in designing public safety-net programs is the inherent tension between broad coverage and incentives to move toward economic selfsufficiency. Broader coverage and more generous benefits provide more assistance in the short term to needy families and individuals, but this creates incentives for persons to utilize government assistance rather than moving into employment. Over the past fifteen years, the United States has increased the incentives for low-income adults to work more, reducing the availability and generosity of benefits for nonworking (and non-disabled) individuals. These policy changes have been connected to substantial increases in work and earnings, particularly among low-income, single-mother families, but they also have made assistance less available to those who find themselves out of work and destitute. The other chapters in this volume discuss topics that are closely related to increased labor-force participation, higher wages, and greater levels of self-sufficiency among less-skilled workers. Unfortunately, there are persons for whom self-sufficiency is not attainable, at least in the short run, given the other constraints in their lives. This chapter focuses particularly on the population of single mothers who report periods without earnings or public assistance income. We refer to these women as “disconnected.” This is a group who had greater access to transfer income in the past, particularly welfare support. With the movement to enforce work, however, these women are not receiving welfare support but have not been successful at transitioning into employment. As we show, this group is very poor. This is not a small population. In our data, 40 percent of single mothers with income below 200 percent of the poverty line report being disconnected for one to four months during a three-year period between 2001 and 2003. Over 18 percent

/

227

Making the Work-Based Safety Net Work Better

report being disconnected for longer than twelve months. Policy efforts to expand safety-net programs to help these women will conflict with efforts to create work incentives. The first section of this chapter looks at shifts in economic need among all persons, particularly among those below the poverty line, for whom safety-net programs might be most important. The second section focuses on defining and describing disconnected single mothers. Over the past ten to fifteen years, we show significant growth in the number of disconnected single mothers who report that they are neither working nor receiving means-tested cash or disability support. This trend is visible across various definitions of disconnection and in both of the national data sets that we use. The third section describes the characteristics of these women. The fourth section investigates the frequency with which single mothers experience a period of disconnectedness and the length of these periods without earnings or transfer support. The fifth section looks at the availability of alternative sources of support for these women. The sixth section discusses the pros and cons of possible policy responses that would strengthen safety-net support for this population. The last section concludes.

CHANGES IN POVERTY STATUS AND ECONOMIC NEED Safety-net programs are particularly important to families in extreme economic need. The greater the share of families with very low income levels, the greater the concern about an adequate safety net. A group of particular concern is single mothers. The welfare reforms of the mid-1990s significantly decreased the availability of cash assistance to low-income families with children, primarily affecting poor single mothers. Mothers were given incentives to move into work and also faced mandates to participate in welfare-to-work programs. The result was a large decline in welfare participation and a significant increase in earnings among these families (Blank 2006). Given these major policy changes, the ongoing well-being of this group is particularly important to understand. Figure 7.1a shows the share of people living in families whose income falls within a specific range relative to the official poverty line in the years 1990, 1995, 2000, and 2005. In 2005, 5.4 percent of all persons lived in a family that reported income below 50 percent of the poverty line; we will refer to this group as “extremely poor.” For a mother with two children, this implies annual cash income of less than $7,900 in 2005; for a single individual, this implies annual cash income of less than $5,100. In 2005, 12.6 percent of the population had income below the poverty line, and 31 percent were poor or near-poor, with incomes below 200 percent of the poverty line. The numbers show variation over the business cycle, with rising numbers of extremely poor and poor individuals during the short recessions of the early 1990s and 2000s, and decreased shares in all poverty groups during the period of sustained economic growth and low unemployment in the late 1990s.

228

/

FIGURE 7.1a

/ All Persons, by Family Income as a Percentage of Poverty-Line Income, 1990 to 2005

10.0

9.2

9.7

8.0 6.8

6.0

9.6

10.1 9.1

8.9 8.9

8.3 8.5

9.5

7.2

5.4

5.2 5.3 4.5

4.0

< 50

51 to 100 100 to 150 Percentage of Poverty–Line Income

2005

2000

1995

1990

2005

2000

1995

1990

2005

2000

1995

1990

2005

1990

0.0

2000

2.0

1995

Percentage of Population

12.0

150 to 200

Source: Authors’ tabulations based on Current Population Survey, Annual Social and Economic Supplement, various years.

FIGURE 7.1b

/ Persons in Single-Mother-Headed-Families, by Family Income as a Percentage of Poverty-Line Income, 1990 to 2005

30.0 24.0

23.5 20.5

24.0

19.4

20.0

18.5 18.3

16.8

16.7

15.5

15.8

13.8

15.0

13.0 13.3 11.0 11.4

10.0

< 50

51 to 100 100 to 150 Percentage of Poverty–Line Income

2005

2000

1995

1990

2005

2000

1995

1990

2005

2000

1995

1990

2005

2000

0.0

1995

5.0

1990

Percentage of Persons

25.0

150 to 200

Source: Authors’ tabulations based on Current Population Survey, Annual Social and Economic Supplement, various years.

/

229

Making the Work-Based Safety Net Work Better

Figure 7.1b shows similar numbers for persons in families headed by a single woman age eighteen to fifty-four with at least one child age eighteen or younger. Compared to the overall population, a far higher share of those in single-mother families is poor or near-poor, with 66.8 percent in this situation in 2005. A shockingly high 24 percent of persons in single-mother families were extremely poor in 1990; the share in extreme poverty fell substantially, however, down to 16.7 percent by 2000. The share of persons in single-mother families in overall poverty fell five percentage points from 1990 to 2000. Although overall poverty changed little from 2000 to 2005, extreme poverty rose by 2.7 percentage points between these two years, suggesting that the women counted among the poor in 2005 had lower average incomes than in 2000. These income changes among single-mother families are particularly striking, as they were not observed for other groups, such as married couples. As overall poverty fell in the 1990s, the share of single-mother families between 100 and 200 percent of the poverty line increased somewhat, from 24.8 percent in 1990 to 29.8 percent in 2000. This suggests a shift among some single mothers out of poverty and into the near-poor category. There are a variety of reasons to focus on changes in economic need among single-mother families. These families include children, and high rates of poverty among these families are a primary reason for high children’s poverty rates in the United States. In fact, 23 percent of all children and 69.3 percent of extremely poor children lived in a single-mother family in 2005. As figure 7.1b indicates, there have been significant shifts in economic well-being among this group. Although average incomes have risen, as we will see, there is evidence that a growing number of women are both off welfare and not working. This is a group for whom questions about safety-net support might be particularly acute. Table 7.1 looks at the changing composition of income between 1990 and 2005 among single-mother families where the mother has less than a high school education; this group is highly likely to be poor. Note that here, as in all tables in the chapter, we calculate poverty status using the census definition of income sharing. This means that we assume that all related persons who live together share income and are part of the same family. This is particularly important for single mothers, since many of them live with other relatives. In 2005, 17 percent of single mothers in poor or near-poor families (whose income was below 200 percent of the poverty line) lived with relatives. Hence, as seen in the fourth column of table 7.1, more than a quarter of income in these families comes from other adults. The data in table 7.1 indicate that average income (in inflation-adjusted dollars) rose steeply among less-skilled single mothers between 1995 and 2000, with very strong growth in own earnings more than offsetting a substantial decline in public assistance support. By 2005, the majority of income among these single mothers came from their own earnings, and the contribution of public assistance income fell to only 2 percent of their total income. This growth in earnings reflects the surge in labor-force participation among less-skilled single mothers, which has been linked to a variety of policy reforms as well as the strong economic growth during this period (Blank 2002; Grogger and Karoly 2005). After 2000 these in-

230

/

Disconnected Single Mothers TABLE 7.1

/ Income Components for Families Headed by Single Mothers with Less than a High School Education

Income Components (Percentages)

Year

Average Total Family Incomea

Family PublicAssistance Income

1990 1995 2000 2005

$22,022 $23,891 $27,002 $25,023

10.3 8.1 2.4 2.1

Own Earnings

Other Family Members’ Earnings

Other Family Income

44.9 44.8 52.7 53.2

28.2 27.8 28.1 26.7

16.5 19.4 16.8 18.0

Source: Authors’ calculations based on Current Population Survey data, Annual Social and Economic Supplement, various years. a. All monetary values in real-year 2005 dollars, deflated using the BEA’s PCE price deflator.

come gains fell off somewhat, but in 2005 average income among less-skilled single mothers was substantially higher than in 1990. Changes in earnings by other family members or changes in other income sources were relatively minor over this time period, although these other sources of income constitute a very high share of the resources available to single mothers and their children. It is worth noting that these data, based on reported cash income, are subject to a variety of caveats and miss some important forms of support. In particular, inkind resources, often available through public assistance programs such as Food Stamps or Housing Assistance, are not counted. On the other hand, the evidence is quite mixed as to whether those most in need are the ones who receive program benefits; this seems to vary across populations and programs. As Janet Currie (2006) notes in her summary of the literature on program take-up, programs that target populations that have difficulty dealing with complex eligibility requirements (such as the elderly or disabled) may find it hard to get benefits to the most needy. There is also a debate about whether the data on extremely poor families are accurately reported. Bruce D. Meyer and James X. Sullivan (2006) have noted that reported consumption among the extremely poor seems higher than reported income. This suggests that there are measurement problems and these families are underreporting their actual income, or that these families are able to draw down savings or build debt in order to smooth their consumption. We are willing to believe that these families have more resources than they report; indeed, it is hard to understand how they survive if this were not the case. But even allowing for a doubling of incomes among the extremely poor still leaves them below the poverty line. As already mentioned, we follow standard practice in defining poverty and assume that only related persons who live together share income. This means assuming that male and female cohabiters do not share income. Previous research

/

231

Making the Work-Based Safety Net Work Better

suggests that resource-sharing among cohabiters is limited. Catherine Kenney (2003) and Anne E. Winkler (1997) both indicate that cohabiting couples share much less income than do married couples. Kenney (2003) and Kurt Bauman (1999) both conclude that cohabitation does much less to reduce economic hardship than marriage. Bauman looks explicitly at the question of whether cohabiters’ income should be counted as shared for the purpose of poverty measurement and concludes that it should not; Kenney (2003) reaches the opposite conclusion. In calculating poverty rates, we follow the standard approach of assuming that only those who are co-residing and related are pooling income. The share of low-income single mothers who are living with an unrelated male has remained quite stable over the 2000s, at about 18 percent, so factors other than changes in cohabitation behavior drive the results we will discuss. (To test the sensitivity of our results to this assumption, when we look at disconnected women in the analysis that is to come we investigate how our results vary if we exclude mothers who live with men who have substantial earnings.) As seen in figure 7.1b and table 7.1, the share of single mothers who report very low incomes increased between 2000 and 2005, and the share of public assistance in single mothers’ income has fallen steadily since 1990. Even if the levels may be underreported, this trend is worrisome. It is consistent with other evidence indicating that some women who left welfare have not found stable employment. For instance, Lesley J. Turner, Sheldon Danziger, and Kristin S. Seefeldt (2006) follow a cohort of female welfare recipients for six years, starting in 1997. They find that 9 percent become “chronically disconnected,” without employment or cash welfare, during at least one-fourth of the seventy-nine months they are observed. Gregory Acs and Pamela Loprest (2004) indicate that a third of those on TANF in 1997 are neither working nor on welfare when they are interviewed in 1999. Rebecca Blank (2007) provides evidence that the number of women without earnings or cash public assistance has grown over the 2000s.

DEFINING AND MEASURING DISCONNECTED SINGLE MOTHERS Historically, welfare assistance has provided a safety net of support for single mothers who were not employed. Given the recent policy emphasis on moving women off welfare and into work, we want to identify those for whom these policy efforts might have created hardship—that is, women who have left welfare but who have not moved into the labor market. In this section, we look at a variety of ways to define this group of women. We use two different data sets throughout this chapter, both of which collect information from a random sample of the entire U.S. population. In both data sets we focus on the population of single mothers age eighteen to fifty-four whose overall income is below 200 percent of the official poverty line and define disconnected single mothers from among this population. The initial results that we show come from the Current Population Survey’s (CPS) Annual Social and Eco-

232

/

Disconnected Single Mothers

nomic Supplement, which asks questions about income sources and employment over the past year. The CPS includes a relatively large sample each year; we have 3,532 single mothers whose family income is below 200 percent of the poverty line (the group we are most interested in) in 1990, 4,801 such observations in 2000, and 4,781 such observation in 2005. Each of these samples is based on a cross-section of women for that year, providing a useful look at how the composition of this population is changing over time. The Survey of Income and Program Participation (SIPP), by contrast, collects data through a series of longitudinal panels, each of which follows a population sample over time. We use the panels that started in 1990, 1996, and 2001. The 1990 panel runs for thirty-two months, the 1996 panel runs for forty-eight months, and the 2001 panel runs for thirty-six months. The SIPP panels follow the same people over time; participants are interviewed every four months. Although the SIPP sample is smaller than that of the CPS, it contains a similar number of individuals who are ever low-income single mothers in a given panel. This similarity in counts is due to the oversampling of low-income individuals, the multiple-year time span of the SIPP panels, and the fact that income levels are measured with monthly frequency in the SIPP. There are 3,045 women who are ever single mothers with family income below 200 percent of the poverty line in the 1990 SIPP, 5,380 such women in the 1996 SIPP, and 4,272 such women in the 2001 SIPP. SIPP asks about income and earnings data at a monthly level, so the SIPP definitions refer to information on last month’s income rather than last year’s. Table 7.2 uses these two data sets to measure trends in the number of disconnected single mothers over time. The top panel is based on calculations from the CPS, and the lower panel uses the SIPP. We provide three sequential definitions of “disconnected,” to make sure that the results are not sensitive to any one definition. Definition 1 identifies women in the CPS who report no earnings or welfare receipt over the entire past year and who do not report their major activity as going to school. (In the lower panel this is based on women in the SIPP who report no earnings or welfare last month.) Definition 2 requires that women report very low levels of earnings and low levels of welfare receipt over the past year (upper panel) or past month (lower panel).1 In both of these first two definitions, “welfare income” is defined as income from either AFDC (Aid to Families with Dependent Children) or TANF (Temporary Assistance for Needy Families), the cash support programs before and after welfare reform. In moving women off welfare, many states pushed to identify families who might qualify for Supplemental Security Income (SSI), which provides cash support to women or children with serious disabilities. If we want to identify women without public assistance, we may want to exclude SSI support as well. Definition 3 is identical to definition 2, but also excludes any women who report that their family received more than $1,000 in SSI income over the year. Although these three definitions result in different shares of women who are defined as disconnected, they are strikingly similar in the trends that they reveal over time. Look first at panel 1, based on CPS data. Using definition 1, the level of disconnected women rises from 10 to 20 percent between 1990 and 2005. Using

/

233

Definition 1a

Definition 1e

Definition 2f 5.4 5.4 7.6 8.7

(1) Living With at Least One “Connected” Adult

5.0 5.7 6.6 8.0

6.4 6.1 8.0 12.0

(3) Living Alone, Without Other Adults

4.0 2.3 2.4 3.1

(2) Living with Other Disconnected Adults

9.5 8.7 10.8 11.5

(3) Living Alone, Without Other Adults

Using Definition 3

1.0 1.2 1.6 1.7

(2) Living with Other Disconnected Adults

13.6 11.0 13.3 14.6

Without Other Adult Income Help: Sum of (2) & (3)

7.4 7.3 9.6 13.7

Without Other Adult Income Help: Sum of (2) & (3)d

Source: Authors’ tabulation based on Current Population Survey, Annual Social and Economic Supplement; and Survey of Income and Program Participation, 1990, 1996, and 2001 panels. a. Not in school, no earnings, no welfare receipt over past year. b. Not in school, annual earnings ≤ $2,000; annual welfare receipt ≤$1,000 (real year 2005 dollars). c. Not in school, annual earnings ≤ $2,000; annual welfare receipt ≤$1,000; annual household SSI receipt ≤$1,000 (real-year 2005 dollars). This is the “standard definition” used in tables 7.3 to 7.7. d.This is the “stricter definition” used in tables 7.3 to 7.7. e. Not in school, no earnings, no welfare receipt over past month. f. Not in school, 12 x monthly earnings ≤ $2,000; 12 x monthly welfare receipt ≤ $1,000 (real year 2005 dollars). g. Not in school, 12 x monthly earnings ≤ $2,000; 12 x monthly welfare receipt ≤ $1,000; 12 x monthly household SSI receipt ≤ $1,000 (real year 2005 dollars).

19.0 16.4 20.8 23.3

Definition 3g

12.4 13.0 16.2 21.7

Definition 3c

(2) Survey of Income and Program Participation 1990 18.8 19.9 1996 17.0 18.8 2001 23.2 24.5 2003 24.9 26.3

Calendar Year

14.1 15.8 18.6 25.1

Definition 2b

(1) Living With at Least One “Connected” Adult

Using Definition 3

/ Trends in the Percent of Disconnected Single Mothers as Percentage of All Single-Mother-Family Heads Whose Family Income Is Below 200 Percent of the Official Poverty Line

(1) Current Population Survey 1990 9.9 1995 11.7 2000 14.6 2005 20.0

Calendar Year

TABLE 7.2

Disconnected Single Mothers

definition 2, the level rises from 14 to 25 percent, and with definition 3 the level rises from 12 to 22 percent. Thus, each of these definitions suggests that over 20 percent (one-fifth) of all lower-income single mothers are disconnected—receiving economic support neither from their own earnings nor from public assistance—by the mid-2000s. Furthermore, this number rises over the period, with a particularly large increase in the share of disconnected single mothers between 2000 and 2005. We use definition 3 throughout the rest of this chapter, referring to it as the “standard definition” in future tables. We think it best represents what most people mean by “disconnected”—women who have very inadequate levels of earnings, welfare, or SSI assistance. But these women may be living with other adults who provide substantial support for them. The last four columns of table 7.2 explore this possibility. Among the 21.7 percent of women who by our standard definition are disconnected in 2005, about one-third of them (8.0 out of 21.7) live with at least one other adult who is either working or on welfare. About half of these women live with an unrelated male who is working, and the others largely live with relatives who work. Fewer than 10 percent (1.7 out of 21.7 percent) of disconnected women are living with other adults who are themselves disconnected (not receiving significant levels of welfare, SSI, or earnings). The majority, about 55 percent (12.0 out of 21.7 percent), live with no other adults. Although there has been some growth between 1990 and 2005 in the share of women who live with a “connected” adult, the most rapid rise is in the share of disconnected women who live alone; this number doubled between 1995 and 2005. This suggests that the data showing a growing number of disconnected women are identifying a population with serious economic needs. Single mothers who live alone and have neither welfare nor earnings have few resources with which to support themselves or their children. In the final column of table 7.2 we provide an alternative and stricter definition of disconnectedness, which is based only on the subset of disconnected women who live with no other adults or with disconnected adults. These are disconnected single mothers who do not appear to rely on other adults for assistance. This number grows from 7.4 percent in 1990 to almost 14 percent in 2005. In future tables we use both our standard definition of disconnectedness (definition 3) as well as what we call our “stricter definition,” based on this more restricted group. As we will see, the conclusions we reach in this chapter are quite similar regardless of which definition is used. As a check on the data from the CPS, part 2 of table 7.2 presents identical calculations using data from SIPP. The data come from the first waves of the 1990, 1996, and 2001 SIPP panels. We also look at the data from the seventh wave of the 2001 SIPP panel, collected in 2003. Because SIPP collects longitudinal data, there are attrition problems; these are likely to be worse among the most disconnected, so the 2003 SIPP data may provide an undercount of this population. Not surprisingly, when the data are based on income last month rather than last year, the share of disconnected women is slightly higher. It is striking, however,

/

235

Making the Work-Based Safety Net Work Better

how similar the SIPP data are to the CPS data. Both show a significant increase in the share of disconnected women between the mid-1990s and the mid-2000s, although the increase in the SIPP is not quite as large as in the CPS.2 Like the CPS, the SIPP also shows a rise in the share of these women who live without other adults (although not as steep as the CPS), and a rise in the share living with a “connected” adult—one with a job or other source of income.

WHO ARE THESE DISCONNECTED WOMEN? The data in table 7.2, along with the other available evidence, confirm that there appears to be a growing number of single-mother families who have lost access to public assistance, but who have not been able to find stable employment. Table 7.3 looks at the characteristics of these women in more detail, using the CPS data. The first column shows the characteristics of all single-mother families below 200 percent of the official poverty line. The second column shows the characteristics of disconnected single mothers, according to our standard definition (definition 3 in table 7.2); the third column uses the stricter definition from the last column of table 7.2, showing only disconnected women who live with no other adults or with other disconnected adults. Low-income single mothers tend to be very disadvantaged, as column 1 indicates. Over half (54.3 percent) are poor, 64 percent have only a high school diploma or less, and a very high share are persons of color. In comparison to this group, however, disconnected single mothers are far worse off, regardless of which definition of disconnected we use. Over 80 percent are poor. Their total reported family income is well below $10,000. Although many have argued that these women must be cohabiting and receiving income from a boyfriend, only 19 percent report themselves as living with an unrelated male using our standard definition. Strikingly, more than a fourth report not working for health-related reasons, although we use a definition that excludes women who are receiving public disability payments through SSI. Other studies with richer information about individual characteristics provide more information about the group of women who fail to make a successful transition from welfare into work. In particular, a variety of studies have documented the multiple barriers to work that some single mothers face and correlated these barriers with problems in finding and holding a stable job.3 Women who have difficulty finding work following welfare reform are more likely to have health problems, particularly problems of depression and other forms of mental illness. They are also more likely to be caring for someone with health problems, either a child or another relative. Compared to other women, they are far more likely to report a history of domestic violence or to be living in a situation that involves domestic violence (Riger, Staggs, and Schewe 2004). Finally, they are also more likely to have past or current problems with substance abuse (Metsch and Pollack 2005). Welfare leavers who have trouble maintaining steady employment typically face one or more of these problems. One study finds that 57 percent of welfare

236

/

Disconnected Single Mothers TABLE 7.3

/ Characteristics of All Single Mothers and Disconnected Single Mothers in 2005, by Definition of “Disconnected”

Disconnected Single Mothers

Characteristics Percentage living with parents Percentage living with other relatives Percentage living with an unrelated male Percentage living with an unrelated female Percentage living alone, no other adults Average number of children Average number of preschoolers Average own earnings Average earnings from other family members Average family income Percentage white or other, non-Hispanic Percentage African American, non-Hispanic Percentage Hispanic Percentage with education less than high school Percentage with education exactly high school Percentage with education more than high school Percentage listing “health problems” as reason for not working Percentage poor Percentage immigrants

All Single Mothers with Family Income Below 200 Percent of Poverty Line

Standard Definitiona

Stricter Definitionb

11.0 6.0 18.2 2.3 63.7

15.9 7.4 20.1 2.3 55.3

8.1 2.9 1.9 0.5 87.6

1.9 0.6

1.8 0.7

1.9 0.6

$9,802 $2,437 $16,445

$200 $3,933 $9,459

$196 $702 $7,045

40.7 33.2 21.1

41.9 29.0 23.8

42.4 35.2 17.9

23.8

31.8

27.1

40.1

39.7

39.6

36.1

28.4

33.3

17.4 54.3 9.9

25.7 82.1 12.7

31.6 86.2 9.6

Source: Authors’ tabulations based on Current Population Survey, Annual Social and Economic Supplement 2006. Note: All monetary values in real-year 2005 dollars, deflated using the BEA’s PCE price deflator. a. See definition 3 in table 7.2. This includes all single mothers in families whose total family income is below 200 percent of the official poverty line, who are not in school, and with annual earnings ≤ $2,000, annual welfare receipt ≤ $1,000 and annual SSI receipt ≤ $1,000 (real year 2005 dollars). b. Includes disconnected single mothers living without other adults or living with other disconnected adults.

/

237

Making the Work-Based Safety Net Work Better

leavers who were not employed faced multiple barriers to work, whereas only 17 percent of those who were employed faced multiple barriers (Loprest 2003). One of the most striking pieces of evidence comes from a study by LaDonna Pavetti and Jacqueline Kauff (2006) of women who hit the welfare time limit in a county in Minnesota. A series of detailed personal assessments found that all of them had some combination of cognitive limitations, mental and physical health problems, a lack of community and social networks, and limited management and decisionmaking skills. We have almost no information on these types of barriers in CPS data, but the SIPP data provide somewhat richer information. The SIPP asks more questions about why women are not working in any given month. Furthermore, a special supplementary survey is occasionally given to the family asking about child development. Table 7.4 uses data from the 2001 SIPP to look at the barriers discussed in the literature that can be measured in the SIPP. The first column shows the share of low-income single mothers reporting problems. The second column shows the share of these problems among women who ever experience at least one month of disconnectedness. The third column shows the level of these problems among women who experience multiple months of disconnectedness. The first part of table 7.4 indicates that disconnected women are far more likely to report not working because of pregnancy, childbirth, or child-care responsibilities. Although welfare income in the past was often targeted on single mothers who had very young children, these women are not receiving it. A very high share of disconnected women report they are not working because they have chronic health problems or a physical or mental condition that limits work. These numbers are high, even though we have excluded women in the months when they report receiving significant amounts of disability assistance through SSI. The second part of table 7.4 uses data from the supplemental survey, which asks the mother about her children’s development. Because this is given only twice during the panel, the responses may or may not occur at the same time as a mother reports being disconnected. These childhood disabilities are long-lasting problems, however. Although child developmental disabilities are infrequent among all samples presented in the table, they are more prevalent among the disconnected. The final row in table 7.4 shows the share of women who report none of the barriers listed in the table. Fewer than half (42.8 percent) of all low-income single mothers report no barriers, indicating that many low-income mothers struggle with these problems for themselves and their children. Most of these women hold jobs, however, or receive public assistance income. There are more barriers among women who report any periods of disconnectedness. Only one-quarter report none of these barriers; among those with multiple periods of disconnectedness, only 17 percent report no barriers. All of this evidence suggests that a substantial subset of single mothers—particularly those with health, behavioral, and family limitations—will have difficulty finding and holding stable employment when welfare-to-work policies, time lim-

238

/

Disconnected Single Mothers TABLE 7.4

/ Percentge of Disconnected Low-Income Single Mothers Reporting Various Barriers to Work

Percentage Experiencing Barrier

Barrier to Work Reported at least onceb Child in family age two or younger Report not working due to pregnancy or childbirth Report not working due to caring for children or others Report not working due to chronic health condition or disability Physical or mental worklimiting condition Reported in a one-time supplementary survey Child in family with developmental disability Child in family with physical or mental disability Child in family with mental retardation Child in family with other developmental disability No observed barriers

Of All LowIncome Single Mothers

Of Those Who Report Any Period of Disconnectednessa

Of Those Who Report Multiple Months of Disconnectedness

(1)

(2)

(3)

21.8

17.9

20.3

6.8

9.2

12.6

27.7

41.4

55.9

9.5

12.1

16.2

24.3

23.0

29.7

0.8

0.6

0.6

1.8

2.4

3.1

0.9

0.9

0.8

5.1

5.9

6.5

42.8

25.7

17.0

Source: Authors’ tabulation based on the Survey of Income and Program Participation panel 2001. Note: Includes only individuals observed in all waves. a. See definition 3 in table 7.2. This includes all single mothers in families whose total family income is below 200 percent of the official poverty line, who are not in school and have 12 x monthly earnings ≤ $2,000; 12 x monthly welfare receipt ≤ $1,000; 12 x monthly household SSI receipt ≤ $1,000 (real year 2005 dollars). b. For columns 2 and 3, these are barriers reported during a period of disconnectedness.

/

239

Making the Work-Based Safety Net Work Better

its, or sanctions move them off welfare assistance. As tables 7.2 and 7.3 suggest, disconnected women are very poor; even if their actual income is underreported by 10 or 20 percent, they would remain an extremely poor group. Table 7.4 suggests that many of these women face significant barriers to work, given both their own problems and the problems that their children are experiencing. This section documents the growth in the number of disconnected single mothers, particularly over the past five years. These women are highly economically disadvantaged with very low incomes; the majority live only with their children or with other nonworking adults. All of this suggests that this is a population that might be of particular concern, both because it is growing and because the economic problems facing these families might have long-term negative effects on the children in these families.

THE TIMING OF PERIODS OF DISCONNECTEDNESS The period of time that women spend disconnected is an important factor when considering policy responses. Hence, one may want to look at spells of disconnectedness, that is, the length of time that women go without significant earnings or welfare income. Within limits, the SIPP data allow us to do this. The 2001 SIPP panel interviewed participants every four months for thirty-six months, for a total of nine interview cycles, called waves. In the 2001 SIPP, there are 4,272 women who are low-income single mothers at some point in the panel, each of whom is theoretically at risk of becoming disconnected. We identify spells of disconnectedness as sequential periods when women report themselves as neither working nor on welfare or SSI. Although women are asked about the past four months each time they are interviewed, a disproportionate number of respondents report changes in employment or public program receipt only every four months, at the point where they are interviewed (the so-called seam bias problem). Because of this we look only at the last month of each wave (the actual interview month), and therefore have information on women’s economic status every four months during the period they are in the SIPP survey. The top panel of table 7.5 focuses on spells of disconnectedness observed in the 2001 SIPP. Columns 1 and 2 refer to all women who are ever defined as disconnected on the basis of our standard definition (reporting a month in which they receive little earnings or welfare or SSI income). Column 1 reports the number of spells for which we observe both the beginnings (the spell is non-left-censored) and the endings (the spell is non-right-censored) during the course of the panel. Of course, these are necessarily shorter spells, so column 2 reports the number of spells that we observe starting in the data, but includes right-censored spells as well. As indicated in column 2, there are 1,168 women who experience at least one spell of disconnectedness that is observed to start in the SIPP panel; this is 27.4 percent of all low-income single mothers. Columns 3 and 4 look at the same concepts and are based on a sample of women who are disconnected using our stricter definition, that is, they live without other adults or with disconnected

240

/

Disconnected Single Mothers TABLE 7.5

/ Percentage of Single Mothers (with Family Incomes Below 200 Percent of the Poverty Line) By Length of Time Disconnected

Standard Definitionb

Length of Time

Uncensored

Uncensored and RightCensored

Stricter Definitionc

Uncensored

Uncensored and RightCensored

(1) Spells of Disconnectedness (Based on First Non-Left-Censored Spells)a Less than or equal to 4 months 71.2 65.4 74.4 5 to 8 months 16.6 17.8 15.0 9 to 12 months 7.7 7.5 6.6 13 to 16 months 2.9 4.9 3.0 17 to 20 months 0.8 1.5 0.7 21 to 24 months 0.5 1.3 0.2 25 to 28 months 0.3 0.9 0.2 28 to 32 months 0.0 0.7 0.0 Average number of months disconnected Total number of non-leftcensored spells Percentage of low-income single mothers with a non-leftcensored spell

4.4

5.3

4.2

4.8

1168

812

27.4

19.0

(2) Total Time Spent Disconnected (Without Regard to Continuity of Spells) Less than or equal to 4 months 51.7 5 to 8 months 20.1 9 to 12 months 9.8 13 to 16 months 6.9 17 to 20 months 3.7 21 to 24 months 2.9 25 to 28 months 2.0 28 to 32 months 1.5 32 to 36 months 1.4 Average number of months disconnected Total number of individuals ever disconnected Percentage of low-income single mothers ever disconnected

70.4 15.4 6.4 4.8 1.1 0.7 0.6 0.5

56.1 18.1 8.3 7.5 3.6 2.6 1.4 1.2 1.2

7.4

6.9

1,726

1,139

40.4

26.7

Source: Authors’ tabulation based on Survey of Income and Program Participation panel 2001. a. This panel does not include left-censored spells, thus omitting spells that are both right- and left-censored. There were forty-nine both-censored spells that lasted more than twelve months based on the standard definition of disconnectedness and twenty-six both-censored spells based on the stricter definition. b. See definition 3 in table 7.2. This includes all single mothers in families whose total family income is below 200 percent of the official poverty line, who are not in school and have 12 x monthly earnings ≤ $2,000; 12 x monthly welfare receipt ≤ $1,000; 12 x monthly household SSI receipt ≤ $1,000 (real year 2005 dollars). c. Includes only disconnected single mothers living without other adults or living with other disconnected adults.

/

241

Making the Work-Based Safety Net Work Better

adults. Among low-income single mothers, 19 percent experience a spell of disconnectedness, using this definition. The data suggest that a high share of the disconnected spells last for four months or less; depending upon the definition of disconnected, between 71 and 74 percent of uncensored spells are relatively short. This share is only slightly lower when right-censored spells are included. Only a small share of these spells last longer than twelve months, between 4 and 9 percent across the columns in panel 1. A potential problem with these tabulations is that they undercount long spells, since they show only spells that start in the data. If there are a number of very long spells, they will likely be left-censored, and will not appear in these tabulations. The number of spells that are both right- and left-censored provides a rough measure of how prevalent these long spells may be. The notes to table 7.5 indicate that forty-nine spells (among all disconnected women) last more than twelve months but are censored on both ends. If we add the both-censored spells that last longer than twelve months to the non-left-censored spells, it turns out that 13 percent of these spells last twelve months or longer among disconnected women (among those without other adult income help, the percentage is 10.6 percent). We have seen that all disconnected women appear to have very low incomes; women with extremely long spells of disconnection are the most destitute. By focusing on spells, the top part of table 7.5 investigates periods of continuous disconnectedness. If women move in and out of unstable employment, however, one might be less concerned with any particular spell of disconnectedness and more concerned with the total number of months in which a woman reports herself disconnected. Panel 2 of table 7.5 reports the number of months in which a single mother reports herself as without substantial earnings or public-assistance support, without attention to whether these months are connected into a continuous spell. The two columns in panel 2 show this analysis for both of our definitions of disconnected. In the bottom panel, any woman who reports even a single month of disconnectedness is included, without attention to issues of left or right censoring. Hence, 40.4 percent of all low-income single mothers report at least one month spent disconnected, in the first column of panel 2, while over one-quarter of women report at least one month, according to our stricter definition of disconnection in the second column of panel 2. The top part of table 7.5 suggested that most women are in relatively short spells of disconnectedness at any point in time, and the bottom panel of table 7.5 suggests that many women experience multiple spells of disconnectedness. Although less than 10 percent of uncensored or right-censored spells last longer than twelve months, more than double this share of women experience more than twelve months of disconnectedness. Table 7.6 provides information on the reasons why women enter and leave disconnected spells. Part 1 shows reasons for the beginning of disconnected spells, on the basis of first spell observed (if any) for each low-income single mother in the SIPP data set. Part 2 shows reasons for the ending of disconnected spells, on the basis of the first spell-ending observed (if any) for each single mother.4 We base these calculations on our standard definition of disconnection.

242

/

Disconnected Single Mothers TABLE 7.6

/ Reasons for Entering and Leaving a Spell of Disconnectednessa Among Disconnected Single Mothers Age Eighteen to Fifty-five with Family Incomes Below 200 Percent of the Povery Line

First Non-LeftCensored Spell (1) Percent breakdown of Reasons for Starting a Spell of Disconnectednessb Left marriage Child under eighteen entered family Mother aged into sample (became eighteen) Welfare income fell below $1,000 per year or SSI income fell below $1,000 per yearc Earnings fell below $2,000 per yearc Family income fell below 200 percent of poverty line No longer in school as primary activity

6.0% 8.7 1.6 13.8 57.5 5.4 6.8

First Non-RightCensored Spell (2) Percent breakdown of Reasons for Ending a Spell of Disconnectednessb Entered marriage Mother aged out of sample (became fifty-six) No more children under eighteen in family Earnings rose above $2,000 per yearc Family income rose above 200 percent of poverty line Welfare income rose above $1,000 per year or SSI income fell below $1,000 peryearc Entered school as primary activity

5.7% 0.4 8.2 55.1 8.2 15.1 7.3

Source: Authors’ tabulation based on Survey of Income and Program Participation panel 2001. a. Based on definition 3 in table 7.2. This includes all single mothers in families whose total family income is below 200 percent of the official poverty line, who are not in school and have 12 x monthly earnings ≤ $2,000; 12 x monthly welfare receipt ≤ $1,000; 12 x monthly household SSI receipt ≤ $1,000 (real year 2005 dollars). b. Reasons sum to 100 percent (up to rounding error) and are tabulated sequentially, so (for instance) changes in marital status take precedence over changes in earnings. As a result, ordering of reasons matters, although changes in order produce little change in relative magnitudes. c. In real year 2005 dollars, deflated using the BEA’s PCE price deflator.

In part 1, 16.3 percent of the spells start because a woman’s marriage breaks up, a child is born, or she reaches her eighteenth birthday, having previously met all other criteria for being classified as a disconnected single mother. Only 13.8 percent of the spells start because of the loss of welfare or SSI income. (This number would surely have been much higher in the late 1990s, when many women were leaving welfare.) More than half (57.5 percent) of spells start because of a change in earnings, probably caused by the loss of a job. Part 2 shows equivalent reasons for disconnected spells to end. The beginnings of spells mirror endings very

/

243

Making the Work-Based Safety Net Work Better

closely. Fourteen percent end a spell because they get married or they no longer have a younger child. About 15 percent leave because their welfare or SSI income increases, while 55 percent leave because of an increase in own earnings. In short, disconnected spells are much more likely to both begin and end due to a change in earnings rather than by leaving or reentering the welfare system. By the early 2000s, job gains and losses are the primary reason why women are becoming disconnected. The safety net, through welfare or SSI, plays a relatively minor role in helping women escape from (or enter into) a period of disconnectedness. Past history with the TANF program (such as sanctions) could be one reason why women are not seeking welfare while disconnected. These findings yield important lessons for safety-net policies directed at disconnected, single-mother families. Most spells of disconnectedness last eight months or less, and generally begin and end with a shock to the woman’s earnings. This is perhaps not surprising in a post-welfare-reform world. When welfare is less available to single mothers, their economic fortunes rise and fall with their labormarket opportunities. Thus, the most appropriate programs will provide potentially recurring short-term income support rather than a single long period of support, and will focus on helping disconnected women find and maintain stable employment.

OTHER SOURCES OF SUPPORT AVAILABLE TO DISCONNECTED WOMEN The cash resources reported in table 7.1 are not the only resources available to women. They may receive in-kind government support, through Food Stamps, Medicaid, or other programs; they may receive help from nongovernmental organizations through food pantries or community-based service organizations; or they may receive in-kind help from other family members with whom they do not live. Officially, if they received cash gifts from others they should be reporting this to the CPS or to the SIPP; most people believe that in reality, such transfers across families are largely unreported. Table 7.1 indicates that cash support through the public assistance system has fallen substantially, but other programs have not contracted. Historically, many women on welfare were connected with other government programs by their welfare caseworker. As a result, when welfare programs were redesigned to move women from welfare into employment, women’s use of some other government programs decreased. For example, for a few years Food Stamp usage declined almost as fast as welfare usage, even though many women leaving welfare remained eligible for Food Stamps. As a result, the Food Stamp Program worked hard to rework its eligibility rules and outreach so that Food Stamps were more easily used by working low-income adults. Office hours were changed to accommodate working applicants who were only available for evening or weekend visits; longer eligibility certification periods were offered to some eligible recipients; and asset limits on cars (a necessity for many workers) were raised. The result of

244

/

Disconnected Single Mothers TABLE 7.7

/ Percentage of Low-Income Disconnected and Not Disconnected Single Mothers Receiving Public Assistance and Insurance Coverage in 2005

Not Disconnected Type of Assistance or Coveragef Welfare SSI Food Stamps Medicaid Employer-provided group health insurance Any of the above

Disconnected

With With With Significant Significant Significant Standard Stricter Welfare Incomea Earningsb SSI Incomec Definitiond Definitione 100.0% 11.8 92.5 100.0

8.3% 2.2 34.8 57.6

25.3% 100.0 74.9 99.3

4.0% 1.1 44.7 65.4

4.8% 1.2 52.5 65.2

5.7 100.0

37.2 84.0

5.2 100.0

5.6 74.4

5.2 75.5

Source: Authors’ tabulation based on the Current Population Survey, Annual Social and Economic Supplement 2006. a. Annual welfare receipt > $1,000 (real year 2005 dollars). b. Annual earnings > $2,000 (real year 2005 dollars). c. Annual household SSI income > $1,000 (real year 2005 dollars). d. See definition 3 in table 7. 2. This includes all single mothers in families whose total family income is below 200 percent of the official poverty line, who are not in school and have annual earnings ≤ $2,000; annual welfare receipt ≤ $1,000; annual household SSI receipt ≤ $1,000 (real year 2005 dollars). e. Includes only disconnected single mothers living without other adults or living with other disconnected adults. f. Survey responses indicating receipt or coverage are based on whether anyone in the family is covered by the relevant program.

these changes was a turnaround in Food Stamp usage. The earlier declines were reversed and a growing share of working families (particularly single mothers) received Food Stamps.5 Similar efforts were made to cover women and their children with Medicaid. Table 7.7 tabulates the extent to which low-income single mothers participate in assistance and insurance programs in 2005, on the basis of the relatively large sample of such women observed in CPS data. The first three columns show program usage among connected low-income single mothers, divided between those who report high welfare income, high earnings, or high SSI income (a few women may be in multiple categories). The last two columns show program usage among disconnected women, using both the standard and the stricter definition.6 As expected, many more women among the “not disconnected” report welfare or SSI income (by definition, only those with very low levels can be included in the disconnected columns). Among women with welfare or SSI income, a high share also receives Food Stamps and Medicaid. (Table 7.7 is based on data indicating that someone in the family is covered by these programs. We do not know whether all family members are covered. So, for instance, Medicaid coverage is likely to be more available to children in these families than to the mothers.) Single mothers’ connections with public assistance assure their coverage by a range

/

245

Making the Work-Based Safety Net Work Better

of programs. Not surprisingly, few of these women receive private insurance. Even among women with significant earnings, over one-third receive Food Stamps and 58 percent report themselves to be covered by Medicaid. Another 37 percent report having employer-provided insurance. Disconnected women receive less protection from these programs. Their very low incomes suggest that virtually all of these women should be eligible for Food Stamps, yet only about half receive them.7 A larger share (two-thirds) report that someone in the family is receiving Medicaid. This is perhaps not surprising, given the evidence in table 7.4 of substantial own and children’s health problems. Very few of them have employer-provided insurance. The evidence in table 7.7 may be reassuring, since it suggests that a significant number of these disconnected women and their families are not entirely outside the public safety net. A number of them receive Food Stamps and even more receive some help through Medicaid. On the other hand, despite the extremely low incomes of these families, a significant number are not receiving any help from public programs, or are only receiving Medicaid assistance, which provides no help in paying rent and grocery bills. Table 7.4 indicated that a high share of these women report health problems that prevent them from working, yet they are not receiving SSI. In short, many of these families appear in need of greater public assistance than they are currently receiving. Not all assistance comes through government programs, of course. Private organizations also provide support for poor families. Many communities have organizations that run food pantries or soup kitchens, or that provide free access to used clothing. Laura Tiehen (2002) indicates that 6.4 percent of single-parent families report using a local food pantry in 2000. This number is down slightly from 1996. Despite the fact that numerous visits to food pantries are reported over the year, few people can rely on them as a primary source of food assistance. Most food pantries have rules about how often a family can receive help, and help is typically limited to a certain quantity of items. Most evidence suggests that food pantries are used occasionally as a supplement to other resources. In fact, at least one-third of food-pantry users also receive Food Stamps, but visit the pantry toward the end of the month when Food Stamps run low (Daponte and Bade 2006; Biggerstaff, Morris, and Nichols-Casebolt 2002). Other studies of welfare leavers have found that food assistance from community organizations is as common among those who stay on welfare as it is among those who leave (Jensen et al. 2002; Dunton and Mosley 2000). For families in economic need, food pantries are more likely to be available in their community than other types of private help.8 Kathryn Edin and Laura Lein (1997), interviewing poor women in inner-city neighborhoods, find that 22 percent of working single mothers report some income from a community agency during the year, receiving on average $165 in assistance from these organizations. We have no data to indicate how much disconnected women make use of food pantries or other agency help, but we can assume that they are likely to use them at least as much as other low-income single mothers and probably more. Our general reading of the evidence is that food pantries or other community

246

/

Disconnected Single Mothers

service organizations can provide, at best, only limited support to disconnected women. An alternative source of support is other family members, who might provide assistance to relatives in need. Our data already take into account the income available from other related adults who share a residence with the single mother, since we (like the designers of the census questionnaire) assume that all co-resident related individuals share income. The CPS also asks about cash gifts from other (non-co-resident) family members. The amount reported is quite small, but there is reason to believe that such transfers might be underreported (Haider and McGarry 2006). Given this, what other evidence exists about whether low-income single mothers are likely to receive support from non-co-resident family members? There are relatively few studies of kinship support among poor single-mother families, and few of these distinguish between co-resident kin and other kin. Katherine Magnuson and Timothy Smeeding (2005) do make this distinction. They indicate that two-thirds of single mothers received no cash support at all from their families after a child is born; among those who do receive family help, much of it comes through shared expenses due to co-residence. Almost no financial support comes from non-co-resident kin. In an older study, William L. Parrish, Lingxin Hao, and Dennis P. Hogan (1991) indicate that the primary benefit that single mothers receive from nearby relatives is child-care assistance; they provide little financial support. Of course, for disconnected mothers who are not working, child-care assistance is not needed as much as it is for working single moms. Steven J. Haider and Kathleen McGarry (2006) find that only 0.3 percent of income among less-skilled single mothers comes from private financial transfers. Other studies that don’t distinguish support from co-resident and non-co-resident kin also indicate that kinship financial help is limited. Meejung Chin (2002) finds that 30 percent of new unmarried single mothers get financial help from relatives in the year after a child’s birth, but this declines rapidly as the child gets older. Robert F. Schoeni (1997) indicates that lower-income families are much more likely to give time (such as child-care assistance) than money. Edin and Lein (1997) have some of the highest estimates of family support, perhaps reflecting the nature of their inner-city sample. They indicate that 47 percent of working single mothers report some family support, averaging $140 per month among those who receive it (this number includes in-kind as well as cash gifts). Women reported that this support often came with return obligations. Our reading of this literature suggests that outside of shared living expenses, financial support from other non-co-resident relatives is often low for single mothers. Most support comes as child-care assistance from nearby kin, assistance that probably would be less useful for disconnected mothers since they are largely not employed. A final source of financial assistance may come from men who are boyfriends or fathers of a mother’s children. As we have noted, only about 18 percent of disconnected women live with an unrelated male and cohabitors share much less income than do married couples. Nonetheless, these women have potential access

/

247

Making the Work-Based Safety Net Work Better

to the earnings of another adult, one reason why our tables look at disconnected women who live only with their children and other nonworking adults. Nonresident fathers may also be a source of assistance. Information on formal child support payments received by the mother is collected in the CPS and included in our data on financial resources; information on regular cash support outside of formal child support is also available (although it may be underreported). Covert or informal support amounts are relatively low and hard to collect information about. Edin and Lein (1997) report that 42 percent of working single mothers receive income from absent fathers, but they do not break this down between formal and covert support. This constitutes 10 percent of the budget for these women. This number may be high, however, since it was collected in the early 1990s when a more select group of single mothers worked; as more women have left welfare and entered the labor market, these women are less skilled and the fathers of their children are also likely to be lower earners with less income to share. Child support collections have been rising, which would increase formal child support payments, but may decrease informal payments. (We know of no recent research that provides more current information on the budget share coming from absent fathers among working single mothers.) Overall, we know that most of these disconnected women have some resources available to them beyond those that they report to surveyors from the CPS or SIPP. Almost surely these women get help—much of it in-kind—from families, friends, community organizations, boyfriends, and the fathers of their children. Indeed, if they did not get this sort of help, it would not be possible to survive on the incomes they report. We would be very surprised, however, if this other income constituted enough to change our overall conclusion that this is a very poor group whose numbers are growing. Certainly these other sources of income are unreliable and variable, and do not offer the economic security that stable employment or public assistance support would provide.

POSSIBLE POLICY RESPONSES In past decades, increases in the number of poor, unemployed women would be likely to generate a conversation about increasing take-up of welfare among this eligible population. In the current policy environment, many of these women were once on welfare but have been encouraged to leave. The evidence suggests that many of the most disadvantaged women who are neither working nor on welfare have hit time limits or have been sanctioned, making it impossible for them to utilize welfare as an income source (Fording, Schram, and Soss 2006; Pavetti, Derr, and Hesketh 2003). The difficulty of returning these women to welfare programs has increased with the recent federal revisions in the Temporary Assistance to Needy Families (TANF) block grant, which occurred in January 2006. TANF provides the primary federal funding stream for state cash welfare programs. The new law requires that

248

/

Disconnected Single Mothers

50 percent of the current welfare caseload must be working in order for states to receive their TANF funding.9 States faced such requirements in the past, but a legal provision allowed them to reduce caseload work requirements if their caseloads fell after 1995. Since all states experienced sharp caseload declines after 1995, state caseload work requirements were also reduced. The January 2006 legislation “resets the base” to the 2005 caseload levels, requiring 50 percent of the caseload to work, and allowing a reduction in this fraction only if caseloads decline after 2005. Few states currently meet this 50 percent requirement. “Work” includes employment as well as a variety of approved work activities, such as supervised job search or job training programs. Women must take part in employment or work activities for at least thirty hours per week to be counted as “working” (twenty hours if they have a child under age six). The result is that states are increasingly concerned not only with moving women off welfare and into work but also with increasing work hours among current welfare recipients. Providing expanded assistance to disconnected women—women who have already demonstrated difficulty in holding stable employment—is likely to prove expensive and not very politically attractive to many states. One option, of course, is to do nothing and to avoid any policy changes focused on disconnected women. If special resources are provided to those who experience an extended period off work and off welfare, this sends the message to women who have difficulty finding a job that they should “wait it out” until they can get additional assistance. Those who insist on sending a strong message about the value of work are likely to resist efforts to build a better safety net for disconnected women. Furthermore, if many of these women have already faced sanctions or time limits, special assistance overrides the purpose of those policies. “Doing nothing” would be more compelling if we thought that these women could work more steadily and earn higher incomes, but have chosen not to. The evidence we present here or that we summarize from other research suggests that a very high share of these women face serious barriers to work that are not easily dealt with, such as domestic violence, depression, caring for a sick or developmentally disabled child, learning disabilities, or other impediments. Furthermore, we are very worried about the effects of low and unstable family incomes on the children in these families, particularly given the health and developmental problems these children already face. So far as we know, no study to date has looked at child outcomes among women who experience extended spells off both work and welfare. We do know that periods of extreme poverty have long-term negative effects on children’s health and school outcomes (Duncan and Brooks-Gunn 1997). Even if one has mixed feelings about helping the mothers, one may argue for supplementing family income if the harm done to children in these households is great enough. Research is needed on how children in these disconnected families are faring. For all of these reasons, we believe that proactive policy attention needs to be given to the growing problem of disconnected single-mother families. We discuss four options here: expanding in-kind program take-up; expanding SSI eligibility;

/

249

Making the Work-Based Safety Net Work Better

designing new state (or federal) programs aimed specifically at this population; and revising welfare rules.

Option 1 Expand support for noncash, means-tested programs, and work hard to enroll disconnected women in these programs. This option requires greater efforts to increase participation in Food Stamps, housing programs, Medicaid, or other noncash assistance programs. The evidence suggests that about half of disconnected women are already enrolled in Food Stamps (see table 7.7), but increasing take-up among the remaining women who are not enrolled would expand their resources. Women who have left welfare—especially those who were sanctioned or timelimited—may be misinformed about their ongoing eligibility for Food Stamps. Similarly, it would be useful to make sure that children from these families participate in school lunch or breakfast programs, and that families are aware of their Medicaid options. Some states are working to keep these families enrolled in Medicaid or Food Stamps, and providing them with special counseling and assistance when their TANF eligibility ends. There may be best practices to be learned from these states that could be more broadly helpful. It may be useful to think about the possibilities for demonstration projects, focusing on generating greater take-up in targeted, in-kind assistance programs among eligible low-income families. Since program officials have made efforts to increase take-up among working single mothers, the implementation of this option is already under way. The problems that prevent women from working can also make it difficult for them to participate in public assistance programs, however. At a minimum, in-kind public assistance programs should think about the particular challenges that may be faced by disconnected single mothers and develop a strategy to encourage greater takeup among those eligible within this population.

Option 2 Expand SSI to allow for partial or temporary coverage. At present, those eligible for SSI must show proof that they are permanently unable to work. As Richard V. Burkhauser and Mary C. Daly (2002) have discussed, U.S. disability programs are open only to those who are permanently disabled, with no part-time or temporary disability coverage. Public programs that provide partial or temporary disability coverage are much more common outside the United States. The advantage of adopting such a program would be that more individuals who face health limitations on work might receive support through SSI. Essentially, this policy change would provide cash support to a larger subset of low-income individuals, through disability payments rather than welfare. Revising the SSI program to include partial or temporary disability determina-

250

/

Disconnected Single Mothers

tion and benefit payments would require a very substantial change, one that would take time and thought to do correctly.10 This type of major reform is often difficult to enact and implement. Any effort to define “disability” is difficult, and the further one moves away from a severe disability eligibility rule, the harder it becomes to establish clear and justifiable eligibility rules. Rising use of disability programs among older workers indicates the extent to which greater eligibility might lead to much higher SSI caseloads and costs. Changes in SSI eligibility would apply broadly to all low-income individuals, not just single mothers. A temporary or partial SSI program is likely to attract far more older persons (too young to collect Social Security and Medicaid, but facing health limitations) than single mothers. Hence, this “fix” to assist disconnected single mothers would likely have many far-reaching consequences. Helping a broader set of adults with health limitations might be a good thing to do, but it is likely to be costly and bring far more people into the SSI disability payment system. Furthermore, revising the SSI system would only help those women whose failure to find stable employment is due to health-related reasons. In general, this option proposes a potentially large and costly reform, which may have a limited impact on the population of single mothers that we most want to help.

Option 3 Create special programs aimed specifically at helping highly disadvantaged single mothers who are having problems finding stable employment. The changes in work requirements for TANF recipients are already generating efforts within some states to create what are known as solely state-funded programs (SSFs), which are entirely outside the TANF structure and whose funding does not count toward the socalled state maintenance-of-effort (MOE) requirements for TANF programs.11 These programs can be used to provide long-term support to women who are not able to find steady employment, or to provide ongoing support to women who do not meet work requirements or who have been removed from TANF support because of sanctions or time limits. The Center on Budget and Policy Priorities (2007) lays out the possibilities for such programs. A number of states are working to screen and identify women with serious barriers to work, including depression or other types of health problems. Programs aimed at “hard-to-employ” women are being tried and evaluated in several locations. So far, however, these efforts are scattered and there is no consensus on what constitutes best practice for these efforts. One proposal for this sort of program is made by Blank (2007), who proposes a Temporary and Partial Work Waiver Program. This would be run by states, allowing them to provide either short-term support—say, to help supplement income for a woman dealing with an ill elderly relative who cannot live alone—or a partial benefit payment—say, to help a woman whose personal problems prevent her from holding a job that pays enough to support her family but who can find and maintain at least some employment. This type of program would require ongoing

/

251

Making the Work-Based Safety Net Work Better

monitoring and more extensive casework than a traditional welfare program. Blank emphasizes the need to regularly reassess women’s ability to work, pushing women whose personal situations change to move more fully toward work when possible. A key goal of an effective program is not only to provide ongoing benefits but also to provide assistance that helps eliminate the barriers to work a woman faces. For instance, such a program could fund substance abuse treatment, counseling, special learning classes, or other assistance to help women overcome barriers to work. These are expensive services and there are often constraints on the number of treatment slots available. Some states may be willing and able to fund more of such services than others. Of course, it would be useful if federal funds were available to supplement state funds for such a program. Federal aid would be dispensed through a program separate from SSFs, and the interaction between this program and TANF would have to be worked out. Federal funds could help low-income states operate such a program and could help smooth the greater cyclicality in state budgets, which cannot provide the sort of deficit financing available to the federal government. The biggest drawback to such a proposal at the state or federal level is the additional administrative machinery and bureaucracy that a new program requires, much of it potentially duplicating some of the casework and assessment activities already found within the TANF program.

Option 4 Remove the barriers within TANF that make it difficult for states to provide ongoing support to single mothers who have difficulty finding and keeping employment. With the recent revisions in the federal TANF block grant, some states are already implementing SSFs. Fifty separate state programs may not be the best way to deal with the problems of single mothers who are unable to find steady employment but who are no longer on welfare. An alternative, of course, is to revise the TANF rules so that some groups of women can receive TANF benefits for a longer period of time, without facing the same work requirements. In the short run, at least, this is not likely to happen. Over the past ten years TANF has become less and less of a safety-net program, and has focused more on moving women into work. Although this has accomplished many good things, it is not obvious that it is a workable solution for all women. Some recognition of the growing share of disconnected women could conceivably result in rolling back some of the recently increased work requirements that states face in their TANF programs, or allowing states to exempt some share of women from those work requirements while still funding them as TANF recipients. Alternatively, states could be encouraged to consider some ongoing assistance to families that face time limits or sanctions; this need not be cash support, but could be in-kind help or help to address personal barriers that limit women’s ability to work. In many ways these changes to the existing TANF program might be more attractive

252

/

Disconnected Single Mothers

than option 3, since they can be done within the parameters of an existing program and do not require creating new program machinery. Even some relatively small changes in TANF might make a difference. For instance, it would be useful to change the federal rules so that TANF does not count months in which a woman on welfare meets work requirements against the TANF time limit. This would allow women who are working part-time to indefinitely continue to receive partial benefits. Another option could be revising MOE requirements to allow states to put more of their own money into SSFs to help disconnected women who are no longer on TANF. The biggest constraint on changes in the current TANF program is political. This program has become known for its push to move women into work and off cash support. Making the sort of revisions suggested here would weaken the nature of these reforms. Whether this is attractive depends upon weighing the value of work incentives for all single mothers against the problems faced by those who have found it difficult to hold steady employment. Before ending this policy section, we want to comment on our focus on singlemother families. There are many extremely poor individuals who do not have children but who suffer economic hardship and do not hold steady jobs. These persons may also benefit from efforts designed to provide greater safety-net coverage to those who have difficulties finding employment. Ideally, a safety net should be there for all those who are destitute. Historically, the United States has never provided much public assistance to non-disabled adults who do not care for children. The value of maintaining strong work incentives for this population has always been viewed as more important. The Food Stamp Program is a notable exception, available to all low-income individuals and families; greater Food Stamp outreach efforts to eligible nonrecipients are likely to benefit more than just single mothers. Other programs are not available to adults who do not care for children. As long as these individuals are viewed as undeserving of public support and expected to work, the political effort to include them in any expanded safety net program is likely to be an uphill battle. Our own advice would be to focus first on the single mothers, a group that citizens have always been more willing to help, and then see if anything learned from that effort that might be useful to apply to other extremely poor families or individuals.

CONCLUSIONS Although many effects of welfare reform have been positive, one disturbing trend is the growing number of disconnected single mothers, who for some period of time go without earnings or welfare support. The recent shift in public assistance programs has left this group without a regular source of public support. Many of these women and their children receive Food Stamps or Medicaid, but many do not. A significant minority of these single mothers report health problems but do not receive disability income. Hence, these families have very low incomes and limited support.

/

253

Making the Work-Based Safety Net Work Better

Our information about this group of women is still quite limited. As previously noted, further research that looks at how these women are actually surviving during their spells of disconnectedness would be very useful. This would also help us judge the extent to which the women and children in these families are experiencing the problems that extreme poverty often brings: residential instability, lack of health care, occasional food shortages, and psychological distress. Given the low reported incomes in these families, the evidence that many of these women face multiple problems that limit their employability, and the needs of the children in these families, it may be foolish to wait for more research evidence before we begin to confront this growing problem. Changes in public assistance programs like those laid out in options 1 through 4 are worth conversation and consideration. Our own preference would be to roll back some of the stricter TANF rules and allow states to use funds to support these women as part of the TANF program or a closely aligned state program. If this is not feasible, providing entirely separate state programs aimed at this population is another option. Finally, ongoing efforts to increase disconnected women’s involvement with other public assistance programs, particularly Food Stamps and Medicaid, are also likely to be helpful. This country has chosen to limit its safety net for poor nonworkers in favor of greater support for those who work. Recent history has demonstrated that many single mothers are able to work, allowing them to receive supplementary support through work-oriented assistance such as the Earned Income Tax Credit. Our concern is for those who have not benefited from these program changes and who have not found steady employment. The preceding analysis demonstrates the serious need for a more effective safety net for these women and their children, warranting an equally serious response by policymakers.

This chapter was originally prepared for the project “Pathways to Self-Sufficiency,” organized by the Institute for Research on Poverty, University of Wisconsin, Madison, September 6 and 7, 2007. Thanks to Carolyn Heinrich, Karl Scholz, Jim Ziliak, and conference participants for useful comments.

NOTES 1. In the CPS data in panel 1 of table 7.2, women must have annual earnings below $2,000 and welfare income below $1,000, and not report “in school” as their primary activity during the past year. In the monthly SIPP data in panel 2, 12 × earnings must be less than $2,000 and 12 × welfare receipt less than $1,000. All income is calculated in inflation-adjusted 2005 dollars. 2. Both the CPS and SIPP appear to be subject to increasing rates of underreporting in TANF income during the period in question, as demonstrated by Laura Wheaton

254

/

Disconnected Single Mothers

3.

4.

5.

6. 7. 8.

9. 10. 11.

(2007). However, the increase in underreporting was focused in the 1990s, while the fraction of disconnected single mothers increased most sharply between 2000 and 2005, implying that the increase in disconnectedness seen in table 7.2 is likely not the result of changes in reporting rates. Key research that presents evidence on the range of barriers that limit women’s employment options include Turner, Danziger and Seefeldt (2006), Ellen Meara and Richard G. Frank (2006), Seefeldt and Sean M. Orzol (2005), Matthew Stagner, Katherine Kortenkamp, and Jane Reardon-Anderson (2002), and Robert Moffitt et al. (2002). These reasons are prioritized, so the reasons sum to 100. This means that women whose marriages break up (reason 1) are coded in this category and do not show up in any other category. Clearly, the order in which these reasons are tabulated will matter; we have experimented with slightly different orderings and the conclusions we describe here do not change. Caroline Danielson and Jacob A. Klerman (2006) and Maria J. Hanratty (2006) both try to estimate the extent to which policy changes have driven the recent increase in Food Stamp caseloads; both studies find significant effects. We cannot distinguish between women who do not receive these programs because they are ineligible and women who are eligible but not receiving benefits. Note that the results here are subject to the same underreporting concerns mentioned in note 2. A national organization, Feeding America (earlier called America’s Second Harvest), helps communities organize the collection of unused food from retailers for redistribution through food banks. Feeding America is currently affiliated with over two hundred food banks, which in turn provide food to food pantries, soup kitchens, and other food programs. For more information, by state and region, see http://feeding america.org. For more details on these new federal requirements, see Center on Budget and Policy Priorities (2007). For a description of the SSI program and its operation, see Daly and Burkhauser (2003). MOE funds are funds that states must spend on the welfare-eligible population in order to be eligible to receive their federal TANF grant.

REFERENCES Acs, Gregory, and Pamela Loprest. 2004. Leaving Welfare: Employment and Well-being of Families That Left Welfare in the Post-Entitlement Era. Kalamazoo, Mich.: W. E. Upjohn Institute. Bauman, Kurt J. 1999. “Shifting Family Definitions: The Effect of Cohabitation and other Nonfamily Household Relationships on Measures of Poverty.” Demography 36(3): 315–25. Biggerstaff, Marilyn A., Patricia McGrath Morris, and Ann Nichols-Casebolt. 2002. “Living on the Edge: Examination of People Attending Food Pantries and Soup Kitchens.” Social Work 47(3): 267–77.

/

255

Making the Work-Based Safety Net Work Better Blank, Rebecca M. 2002. “Evaluating Welfare Reform in the U.S.” Journal of Economic Literature 40(4): 1105–66. ———. 2006. “What Did the 1990s Welfare Reforms Accomplish?” In Public Policy and the Income Distribution, edited by Alan J. Auerbach, David Card, and John M. Quigley. New York: Russell Sage Foundation. ———. 2007. “Improving the Safety Net for Single Mothers Who Face Serious Barriers to Work.” Future of Children 17(2): 183–97. Burkhauser, Richard V., and Mary C. Daly. 2002. “U.S. Disability Policy in a Changing Environment.” Journal of Economic Perspectives 16(1): 213–24. Center on Budget and Policy Priorities. 2007. Implementing the TANF Changes in the Deficit Reduction Act: “Win-Win” Solutions for Families and States. Washington, D.C.: CBPP and Center for Law and Social Policy. Chin, Meejung. 2002. “Longitudinal Patterns of Financial Kin Support and Welfare to Single Mother Families.” Journal of Korean Home Economics Association, English edition 3(1): 37–54. Current Population Survey, Annual Social and Economic Supplement. Various years. Produced for the Bureau of Labor Statistics. Washington, D.C.: U.S. Bureau of the Census. Available at: http://www.census.gov/cps (accessed March 26, 2009). Currie, Janet. 2006. “The Take-Up of Social Benefits.” In Public Policy and the Income Distribution, edited by Alan J. Auerbach, David Card, and John M. Quigley. New York: Russell Sage Foundation. Daly, Mary C., and Richard V. Burkhauser. 2003. “The Supplemental Security Income Program.” In Means-Tested Transfer Programs in the United States, edited by Robert A. Moffitt. Chicago: University of Chicago Press. Danielson, Caroline, and Jacob A. Klerman. 2006. “Why Did the Food Stamp Caseload Decline (and Rise)? Effects of Policies and the Economy.” IRP Discussion Paper 1316-06. Madison: University of Wisconsin, Institute for Research on Poverty. Daponte, Beth Osborne, and Shannon Bade. 2006. “How the Private Food Assistance Network Evolved: Interactions Between Public and Private Responses to Hunger.” Nonprofit and Voluntary Sector Quarterly 35(4): 668–90. Duncan, Greg J., and Jeanne Brooks-Gunn. 1997. Consequences of Growing Up Poor. New York: Russell Sage Foundation. Dunton, Nancy, and Jane Mosley. 2000. “The Use of Community-Based Emergency Assistance Among Kansas City Welfare Leavers.” Paper presented to the National Association for Welfare Research and Statistics Annual Workshop. Scottsdale, Arizona (August 2000). Edin, Kathryn, and Laura Lein. 1997. Making Ends Meet: How Single Mothers Survive Welfare and Low-Wage Work. New York: Russell Sage Foundation. Fording, Richard, Sanford F. Schram, and Joe Soss. 2006. “Devolution, Discretion, and the Effect of Local Political Values on TANF Sanctioning.” Social Service Review 81(2): 285–316. Grogger, Jeffrey, and Lynn A. Karoly. 2005. Welfare Reform: Effects of a Decade of Change. Cambridge, Mass.: Harvard University Press. Haider, Steven J., and Kathleen McGarry. 2006. “Recent Trends in Resource Sharing Among the Poor.” In Working and Poor, edited by Rebecca M. Blank, Sheldon H. Danziger, and Robert F. Schoeni. New York: Russell Sage Foundation. 256

/

Disconnected Single Mothers Hanratty, Maria J. 2006. “Has the Food Stamp Program Become More Accessible? Impacts of Recent Changes in Reporting Requirements and Asset Eligibility Limits.” Journal of Policy Analysis and Management 25(3): 603–21. Jensen, Helen H., Steven Garasky, Cory Wessman, and Sarah M. Nusser. 2002. “A Study of Households in Iowa that Left the Food Stamp Program.” CARD Staff Report 02-SR 97. Ames, Iowa: Center for Agricultural and Rural Development. Kenney, Catherine. 2003. “Hardship in Married and Cohabiting Parent Households: Do Cohabiting Parents Underinvest in Household Public Goods?” Paper presented to the Annual Meeting of the American Economic Association. Washington, D.C. (January 2003). Loprest, Pamela. 2003. “Disconnected Welfare Leavers Face Serious Risks.” No. 7 in series “Snapshots of America’s Families III.” Washington, D.C.: Urban Institute. Available at: http://www.urban.org/url.cfm?ID=310839 (accessed March 25, 2009). Magnuson, Katherine, and Timothy Smeeding. 2005. “Earnings, Transfers, and Living Arrangements in Low-Income Families: Who Pays the Bills?” Paper presented to the National Poverty Center’s Mixed Methods Research on Economic Conditions, Public Policy, and Family and Child Well-Being Conference. Ann Arbor, Michigan (June 27, 2005). Meara, Ellen, and Richard G. Frank. 2006. “Welfare Reform, Work Requirements, and Employment Barriers.” NBER Working Paper 12480. Cambridge, Mass.: National Bureau of Economic Research. Metsch, Lisa R., and Harold Pollack. 2005. “Welfare Reform and Substance Abuse.” Milbank Quarterly 83(1): 65–99. Meyer, Bruce D., and James X. Sullivan. 2006. “Consumption, Income and Material WellBeing After Welfare Reform.” NBER Working Paper 11976. Cambridge, Mass.: National Bureau of Economic Research. Moffitt, Robert, Andrew Cherlin, Linda Burton, Mark King, and Jennifer Roff. 2002. “The Characteristics of Families Remaining on Welfare.” Welfare, Children, and Families Study Policy Brief 02-2. Baltimore: John Hopkins University. Parrish, William L., Lingxin Hao, and Dennis P. Hogan. 1991. “Family Support Networks, Welfare and Work Among Young Mothers.” Journal of Marriage and the Family 53(1): 203–15. Pavetti, LaDonna A., Michelle K. Derr, and Heather Hesketh. 2003. “Review of Sanction Policies and Research Studies.” Report prepared by Mathematica Policy Research, Inc. Washington, D.C.: U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation. Pavetti, LaDonna A., and Jacqueline Kauff. 2006. “When Five Years Is Not Enough: Identifying and Addressing the Needs of Families Nearing the TANF Time Limit in Ramsey County, Minnesota.” In Lessons from the Field. Princeton, N.J.: Mathematica Policy Research. Riger, Stephanie, Susan L. Staggs, and Paul Schewe. 2004. “Intimate Partner Violence as an Obstacle to Employment Among Mothers Affected by Welfare Reform.” Journal of Social Issues 60(4): 801–18. Schoeni, Robert F. 1997. “Private Interhousehold Transfers of Money and Time: New Empirical Evidence.” Review of Income and Wealth 43(4): 423–48. Seefeldt, Kristin S., and Sean M. Orzol. 2005. “Watching the Clock Tick: Factors Associated with TANF Accumulation.” Social Work Research 29(4): 215–99. Stagner, Matthew, Katherine Kortenkamp, and Jane Reardon-Anderson. 2002. “Work, In/

257

Making the Work-Based Safety Net Work Better come, and Well-Being Among Long-Term Welfare Recipients: Findings from a Survey of California’s ‘Precarious Families.’” Assessing the New Federalism Discussion Paper 02-10. Washington, D.C.: Urban Institute. Survey of Income and Program Participation. Various years. Washington, D.C.: U.S. Bureau of the Census, Demographics Survey Division. Available at: http://www.census.gov/ sipp (accessed March 26, 2009). Tiehen, Laura. 2002. “Use of Food Pantries by Households with Children Rose During the Late 1990s.” Food Review 25(3): 44–49. Turner, Lesley J., Sheldon Danziger, and Kristin S. Seefeldt. 2006. “Failing the Transition from Welfare to Work: Women Chronically Disconnected from Employment and Cash Welfare.” Social Science Quarterly 87(2): 227–49. Wheaton, Laura. 2007. “Underreporting of Means-Tested Transfer Programs in the CPS and SIPP.” 2007 Proceedings of the American Statistical Association, Social Statistics Section. CD-ROM. Alexandria, Va.: American Statistical Association: 3622–29. Winkler, Anne E. 1997. “Economic Decision-Making by Cohabitors: Findings Regarding Income Pooling.” Applied Economics 29(8): 1079–90.

258

/

Chapter 8 Beyond the Safety Net: Supporting the Economic Security of Working-Poor Families Marcia K. Meyers and Janet C. Gornick

A

persistently large share of all U.S. families with children is poor or struggling at the margins of economic security. In the United States, as in other rich industrialized countries, market income is the most important factor in families’ economic security. Families in the United States and other countries do not depend exclusively on market income, however. Governments supplement the resources of households throughout the income distribution by providing a “package” of social benefits and employment supports. The United States is similar to other rich industrialized countries in providing a social benefits package, for poor and nonpoor families, that includes social insurance, means-tested transfers, and tax-based assistance. The United States differs from many of the OECD (Organisation for Economic Co-Operation and Development) countries, particularly the countries of northern Europe, in the mix of these policy mechanisms and in the extent to which the value and the type of benefits are stratified by income. In this chapter we consider the structure of the U.S. social and employment benefit package for working families with children and its consequences for their economic security. We argue that the United States currently provides a relatively generous and sustainable social benefits package—through tax- and employment-based benefits—but that much of this support is out of reach for low-income families. Drawing on the experience of other industrialized countries we suggest institutional and structural principles for more comprehensive, inclusive, and sustainable social and employment benefits to support the economic security of U.S. working families. In the following section we define the problem, introduce the concept of a social

/

261

Making the Work-Based Safety Net Work Better

benefits package, and provide a brief overview of our arguments. Then we compare indicators of poverty and of the institutional structure of social and employment benefits in the United States with those in other high-income countries, arguing that higher child poverty rates in the United States result in part from the highly stratified structure of social benefits. We provide a more detailed comparison of policies in three areas—income benefit, paid family leave, and child-care policies—in the United States and Nordic countries; the latter serve as a model for a more comprehensive, integrated, and universalistic approach. To round out our discussion we briefly discuss another, less-comprehensive approach to reducing child poverty that is under development in the United Kingdom. In the concluding section we summarize our arguments for moving beyond a fragmented and incomplete safety net for the poorest families in the United States and recommend design principles for a more inclusive and sustainable social benefits system. We conclude with some final thoughts about unintended consequences and costs.

THE PROBLEM OF WORK AND POVERTY The welfare reforms of the 1990s dramatically reduced the number of American families receiving cash welfare benefits. Work effort among U.S. adults with children increased sharply in the 1990s, particularly among single mothers. There is considerable debate about whether the amount of time that Americans spend working for pay has risen in recent decades (see Figart and Golden 1998 for a review). What is absolutely certain is that the rise in mothers’ employment rates over the past forty years, in both married and single-parent families, means that more families now have all adults in the workforce. Although work effort is substantial in the United States, it has not been enough to assure economic security for all families, even during periods of robust economic performance (see, for example, Mishel, Bernstein, and Allegretto 2007; Blank, Danziger, and Schoeni 2006). The Bureau of Labor Statistics (BLS) (U.S. Department of Labor 2006) estimates that, as of 2004, 5.6 percent of adults who were in the labor force for twenty-seven weeks or more had incomes below the official poverty threshold—a rate that was about the same as that observed in 1987 and again in 1996, on the eve of welfare reform. Working families with children are particularly at risk for income insecurity and poverty. The BLS estimates that among working families, those with children are about four times as likely to be poor as those without. The United States is a high-employment economy, even among OECD countries. Labor-force participation among working-age adults is over 90 percent for men and about 75 percent for women (OECD 2007b). On average in the OECD countries, more than 60 percent of mothers with children age sixteen or younger are in paid employment; in the United States, fully 66 percent work for pay (OECD 2007a). Outside of periods of economic recession and high unemployment, contemporary poverty among families with children in the rich countries

262

/

Beyond the Safety Net

results less from joblessness than from a combination of low hours of employment and low wages among those who work. Only about 33 percent of poor children in OECD countries live in a family with no workers, and even fewer poor children in the United States, less than 15 percent, live with nonworking parents (Whiteford and Adema 2007). Low wages are a particularly important factor in the United States, where nearly one-quarter of workers have earnings that are less than two-thirds of median earnings (OECD 2007b). Governments have a variety of mechanisms for addressing low wages, from macroeconomic and active labor-market policies to human capital investments, minimum-wage laws, and other employment regulations. We leave the question of whether and how the United States can increase wages in the context of the globalized economy to other authors in this volume. In this chapter we consider, given the existence of a large low-wage employment sector, what can government do to help families with low-earning parents achieve greater economic security? We use the term “economic security” in this chapter, rather than the term “selfsufficiency,” because we are interested in the combination of earnings and social benefits that allow families to avoid economic hardship while managing their earning and caregiving responsibilities. All households in the United States and other industrialized countries depend on an aggregate of cash and noncash resources that include earnings, other forms of market income, nonwage employment compensation, and a variety of social benefits (Peattie and Rein 1983; Rainwater, Rein, and Schwartz 1986; Rainwater and Smeeding 2004). Although a subset of these social policies is typically labeled “antipoverty” policy, and another subset is labeled “family” or “work and family reconciliation” policy, contemporary labor-market, family, and gender realities are rending these distinctions moot. Most families with children in the United States and other industrialized countries are now “working families” because the majority of children live in families where all parents are in the workforce and earnings constitute the majority of their income. And the direct costs of childrearing, combined with reductions in employment to care for children, place a substantial share of these families at risk for poverty. In this chapter we consider the role of government, in the United States and some European countries, in providing a “social benefits package” that supports economic security for working families. It is possible to define the social benefits package very broadly to include all social, educational, health, and employment benefits. Other analysts define the social benefits package for families with children more narrowly to include only child-specific benefits, excluding important antipoverty and employment supports for parents. In this chapter we chart a middle ground by focusing on three policy areas at the intersection of supporting parents in their earning and caring roles. Income benefits increase households’ disposable income directly by supplementing or replacing earnings and can increase income security indirectly by creating employment incentives. Paid family leaves provide parents with job-protected breaks in employment for childbirth and other family reasons and replace earnings during those temporary job interruptions.

/

263

Making the Work-Based Safety Net Work Better

Subsidized child-care provisions free parents for employment, reduce the impact of child-care costs for low earners, and prevent the growth of a low-wage child caregiving workforce.1

POVERTY AND THE STRUCTURE OF THE SOCIAL BENEFITS PACKAGE Several decades of comparative research have documented the extent to which the United States trails other rich countries in protecting families with children from poverty. The United States is notable for having a large low-wage employment sector, and pre-tax and pre-transfer poverty rates for children are higher in the United States than in many OECD countries. But the United States stands out, in cross-national comparisons, even more starkly for its failure to reduce marketgenerated poverty. Using data from the Luxembourg Income Study, Lee Rainwater and Timothy Smeeding (2003) find that among fifteen industrialized countries (including the United Kingdom, Canada, Australia, and countries in western and northern Europe), the United States lowers child poverty rates by about 25 percent through taxes and transfers, much less than the reductions of 50 to more than 75 percent observed in most of their comparison countries. Peter Whiteford and Willem Adema (2007) reach similar conclusions, using data from the OECD Income Distribution Study. They estimate the market-income child poverty rate in the United States at about 27 percent in the early 2000s, somewhat higher than the OECD average of 20.5 percent. After all taxes and transfers are considered the U.S. child poverty rate drops by only about four percentage points whereas the OECD average drops eight percentage points to about 12 percent. The OECD averages include countries, such as Mexico and Turkey, that have much lower GDP and higher market poverty rates than the United States. In countries with economies that are more comparable to that of the United States, redistributive policies achieve even more dramatic poverty reductions, for example, from 16 to 3 percent in Sweden, from 12 to 2 percent in Denmark, and from 28 to 7 percent in France.

Level and Structure of Social Spending in the United States and other OECD Countries The relatively weak effect of U.S. tax and transfer policies on child poverty suggests that there is something to learn from the social policy approaches of other OECD countries. The United States is often described as an outlier among industrialized countries, owing to it lower overall spending on social benefits and on child and family benefits in particular. A recent OECD report compares public spending exclusively for families (such as child and family allowances, parental leave payments, and child-care subsidies) in 2003—including spending associated with cash benefits, services, and tax breaks for families—and finds that the United States spent less, as a share of GDP, than seventeen of the twenty-four included

264

/

Beyond the Safety Net

OECD countries (OECD 2007a). Shirley G. Gatenio and Sheila B. Kamerman (2006) estimate that the United States spent only 0.4 percent of its GDP on family benefits in 2001, compared to 2.2 percent in the U.K. and as much as 3 percent in several of the countries of western and northern Europe. In purchasing-parity adjusted dollars, they estimate that this amounted to about $453 per child in the United States compared to over $2,000 in the U.K. and over $3,000 in Austria, Denmark, Finland, France, Luxembourg, Norway, and Sweden. Some analysts argue that the United States actually spends more on social benefits than these measures suggest because the measures do not capture private and tax-based assistance. The United States differs from other rich countries not only in how much we spend on social welfare but in how we spend it (Howard 1997; Hacker 2002; Adema and Ladaique 2005; Garfinkel, Rainwater, and Smeeding 2006; Whiteford and Adema 2007). Contributory social insurance programs, transfers financed with tax revenues, and public services are the most common mechanisms for delivering social and employment benefits in most of the OECD countries. To an increasing extent, high-income countries are also providing benefits indirectly through their tax systems. In addition to the imposition of progressive tax rates and exemption of government benefits from taxation, governments use tax policies to achieve social goals through forgone taxes, or tax expenditures, that result from specialized tax deductions, exemptions, and credits for individual tax filers. These tax benefits offset out-of-pocket expenditures (say, through credits for health insurance and child-care expenditures), equalize income across households (for example, through tax deductions and tax credits for dependents), and target other social goals (such as through deductions for home mortgage interest). Although the use of tax benefits is increasing throughout the OECD countries, the United States remains a leader in the use of the tax system to assist families with children. Spending via tax benefits constitutes a relatively small share of total public spending on families in most OECD countries. Tax expenditures play a more substantial role in a few European countries, including France, Germany, Belgium, the Netherlands, and the United Kingdom. The United States is an outlier, among the OECD countries, in that tax credits and exemptions are the largest single component of public spending on families (Whiteford and Adema 2007). The United States is even more of an outlier in the extent to which social benefits are negotiated and provided through private labor and consumer markets but subsidized with tax dollars. Health care and pensions are the largest and most prominent examples of benefits that are negotiated through employment contracts but heavily regulated by government and heavily subsidized through exemptions from employer and employee payroll and income taxes (Hacker 2002; Garfinkel, Rainwater, and Smeeding 2006; Adema and Ladaique 2005). The United States is the only OECD country that heavily subsidizes employerprovided health benefits, and tax exclusions for employer contributions to health insurance and care is estimated to account for one-third of spending on health insurance—over $80 billion, or 0.8 percent of GDP, in 2001 (Adema and Ladaique 2005).2

/

265

Making the Work-Based Safety Net Work Better

The Distribution of Benefits In comparison to other OECD countries, the United States provides more of the total social benefits package for families through tax benefits for individuals, private employment benefits, and tax subsidies for those private benefits. The distributional consequences of this structure are difficult to observe at the household level because nationally representative data sets do not provide complete information about direct, tax-based, and tax-subsidized private benefits. In one of the few studies that compares the value and structure of assistance across households, Meyers (2007; see also Meyers and Garcia 2006) estimates the social benefits package for a representative sample of New York City households collected through the Columbia University Social Indicators Survey (waves II and III).3 The social benefits package for this exercise included the family-size adjusted estimated value of cash, health insurance, housing, food, and early-childhood education and care benefits received through social insurance programs, through means-tested cash transfers, through in-kind goods and services, and through specialized tax benefits for individuals and employers. When all types of social and employment benefits were considered, the authors estimated that about equal shares of the total social benefits package for all New York City households were provided through each of three mechanisms: universal or broadly inclusive social insurance programs, direct tax benefits and taxsubsidized employment benefits, and targeted means-tested assistance. However, the composition of the benefit package varied sharply by household type. The elderly received most of their benefits through universal programs; working-age households without children received mostly tax-based and tax-subsidized employment benefits; and working-age families with children received most of their total social benefits package—nearly one-half on average—as targeted, meanstested assistance. Meyers (2007) found that the social benefits package for families with children in New York City was even more highly stratified by income. Families with children at different income levels received social benefits packages that were similar in value but fundamentally different in structure, because the progressive distribution of means-tested benefits was nearly offset by the regressive distribution of specialized tax benefits and tax-subsidized private benefits. Families in the bottom two quintiles of earnings received the most generous social benefits package and nearly all of their benefits—an estimated 84 percent for those in the lowest quintile and 70 percent in the second lowest—were provided through targeted, means-tested welfare cash and near-cash transfers (including Food Stamps, Temporary Assistance to Needy Families, and Supplemental Security Income) or through welfare-based in-kind benefits, including health insurance and child care. Tax-based assistance, primarily the Earned Income Tax Credit (EITC), was negligible for families in the lowest earning quintile and accounted for about 10 percent of the social benefits package for those in the second-lowest. Families in the highest quintile of market income received a social benefits package that

266

/

Beyond the Safety Net

equaled about 80 percent of that received by those in the second-lowest quintile. Nearly all of the benefits for high earners (75 percent) were in the form of tax benefits, including those for home ownership, child-care expenditures, and employerprovided health care.

Consequences for Families’ Social Benefits Package and Income Security The structure of the U.S. social benefits package varies substantially with age and family structure. Among families with children, the mix of means-tested welfare and tax benefits also varies dramatically with income. If the value of the total benefit package is progressive, that is, if it provides the greatest assistance to those with the lowest market incomes, does this stratified structure matter for the economic security of low-earning families? We would argue that it does, for at least three reasons. First, the fragmented and income-stratified structure of the U.S. system creates discontinuities and gaps in coverage, particularly for the lowest earners. Outside of public education, none of the major social benefit programs for families in the United States is universal in coverage. Separate, partially overlapping, and poorly coordinated public programs for the poorest families, in combination with taxand employment-based benefits for the more-advantaged, provide duplicate coverage to some families and no coverage at all for others. For families whose benefit package is dominated by means-tested programs, the difficulty and stigma of accessing benefits reduces coverage still further. Participation in tax-based and employer-provided benefit programs, and in universal social insurance programs such as Social Security and Medicare, approach 100 percent of those eligible (Remler and Glied 2003). Participation in means-tested programs for low-income populations is much lower, owing in large part to timeconsuming, invasive, and stigmatizing eligibility tests (Remler and Glied 2003). As Janet C. Currie (2006) observes, these eligibility tests often exclude the very individuals that programs are designed to serve because “poor parents are the least able to negotiate complex administrative requirements” (154). When inadequate funding is combined with incomplete coverage and barriers to take-up, low-earning families rarely get the full social benefits package to which they may be entitled. Sheila Zedlewski and Seth Zimmerman (2007) estimate, for example, that only 7 percent of working families with incomes below poverty receive what they call the “core set of supports for low-income families”: Medicaid and the State Child Health Insurance Program (S-CHIP), Food Stamps, child-care subsidies, and federal and state Earned Income Tax Credits. A second reason to be concerned about the structure of the U.S. social benefits package is that the targeted, means-tested programs that help the lowest-income families are not sustainable over time. Most forms of direct means-tested assistance are explicitly or implicitly time-limited. Programs such as TANF impose explicit time limits on recipients. Other social and employment benefits are nonsus-

/

267

Making the Work-Based Safety Net Work Better

tainable because they are targeted to the lowest-income families and benefits are reduced as earnings increase. Benefit reductions can be steep for those with earned income. For example, Jennifer L. Romich, Jennifer Simmelick, and Stephen D. Holt (2007) and Holt and Romich (2007) demonstrate that for families in the relatively high-benefit state of Wisconsin, a doubling of earned income from $15,000 to $30,000 could net a family as little as $250 more per month after adjusting for the loss of means-tested benefits. By their estimates, the marginal tax rate is 50 percent—or 50 cents out of every extra dollar of earnings—for as many as one out of four Wisconsin tax filers who claim the EITC and use two or three means-tested programs. These families are doing slightly better, in absolute terms, as earnings rise, but they face much higher marginal taxes and disincentives to increase their work effort than other taxpayers. And efforts to increase their earnings will, if successful, guarantee that they will eventually lose important components of their social benefits package. A third and related reason for concern about the structure of United States policies for low-earning families is that there are no institutional bridges between means-tested welfare programs and sustainable forms of assistance that are available to higher-income families through tax- and employment-based systems. Unlike means-tested benefits, the assistance received by middle- and upper-income families is sustainable over time and with increased earnings. The value of most tax benefits actually increases with earnings that increase marginal tax rates. The likelihood of obtaining employer-provided benefits also increases with earnings. For example, the likelihood that an employee has employer-provided health insurance increases from 24 percent, for those in the lowest quintile of wage earners, to over 77 percent for those in the highest quintile (Mishel, Bernstein, and Allegretto 2007) and the share of working parents with any paid sick leave rises from 24 percent for those in the lowest quartile of family income to 60 percent for those in the top quartile (Heyman 2000). These are sustainable benefits for families. But most low-earning families will lose targeted, means-tested benefits long before they earn enough to benefit from nonrefundable tax deductions and exemptions other than the EITC, or to secure employment that provides comparable benefits.

ALTERNATIVE APPROACHES Like all public policies, social policies typically reflect trade-offs and incorporate competing social goals. Most obviously, they reflect the trade-off between benefit generosity and cost, and the willingness of populations to tax themselves to redistribute income, wealth, and risk. Because they can alter incentives for work, saving, and family formation, redistributive policies create other trade-offs, usually characterized as the trade-off between efficiency and equity, at the societal level, and between work and welfare at the individual level. Policies designed to support families with children raise even more complex questions about private and shared responsibilities for raising children, about the social and economic contri-

268

/

Beyond the Safety Net

butions of caregiving work, about the allocation of parental time to market and care work, and about gender equality in these allocations. Comparisons between the United States and the other industrialized countries tell us that the United States has higher rates of income poverty among families with children than most rich countries and does less, through tax and social benefits, to reduce the market-generated poverty of these families. Broad comparisons do not tell us much about the institutional details behind systems that do more to reduce the poverty of families with children, however, or how they address these trade-offs. To learn more about the institutional design of policies that support the economic security of poor families we need to look in more detail at the institutional structure or architecture of policies in the United States and other countries. We concentrate our comparisons on the three policy areas introduced earlier: income benefits that supplement low earnings; paid family leave policies that provide parents with time to care for young children without undue economic penalties; and subsidized child care to free parents for employment and reduce out-of-pocket costs for care. There is no single OECD, European, or even European Union approach to such policies, although there are commonalities across many countries. To simplify our comparisons and draw the sharpest institutional contrasts, in the following sections we compare the structure of income benefit, paid family leave, and childcare policies in the United States to those in the Nordic countries, particularly Sweden and Denmark. Some observers argue that political and demographic differences are too great to make the “Nordic model” relevant to the United States. As the authors of the recent OECD report Babies and Bosses: Reconciling Work and Family Life observe, however, the Nordic countries remain the model for policy developments in the industrialized world: “In family policy during the first decade of the 21st century, the model countries usually held up as deserving of emulation are most often the Nordic countries, especially Denmark and Sweden” (OECD 2001, 73). As the authors go on to note, however, these programs are expensive and may be difficult to sustain in countries with less-cohesive populations and less-egalitarian social principles. Although we believe that the Nordic countries provide the most inspiring and informative models for policy development, we round out this section by briefly discussing a less-comprehensive approach to reducing child poverty that is in development in the United Kingdom.

Income Benefits Income benefits support families’ economic security directly and indirectly by creating incentives (or disincentives) for mothers and fathers to enter employment or to work more hours. The package of income benefits for families with children in most industrialized countries includes employment-related benefits; child and family allowances paid directly or through the tax system; and cash transfers conditioned on income (means-tested) or family composition (oneparent allowances and child support or advance maintenance payments). Income

/

269

Making the Work-Based Safety Net Work Better

benefits throw the trade-offs between generosity and public costs into particularly sharp relief. They also raise questions about the balance between “cash” and “employment” policies.4 Generous income benefits can directly reduce income poverty but, in the absence of employment supports and expectations, transfers that substitute for earnings can lead to long-term employment reductions or withdrawals from the labor force, particularly among women caring for children or other dependents. On the other hand, very low cash benefits and rigorous work requirements can create incentives for employment. When coupled with affordable child care, paid parental leaves, and good wages and benefits, these policies can be said to “make work pay.” When these benefits and protections are absent, they are more accurately described as “work forcing” and risk weakening employees’ bargaining power with employers and, for parents, creating difficult choices between income poverty from reduced work hours and “parental time poverty” from the reduction of time in the home. THE UNITED STATES APPROACH The United States approach to income benefits for families emphasizes minimizing costs, targeting assistance to the most needy, and reducing employment disincentives. The three most important cash assistance systems for families with children—contributory Unemployment Insurance (UI), tax-based child and family benefits, and means-tested cash welfare—reflect these priorities. Although Unemployment Insurance is not usually included in discussions of either antipoverty or family benefits, it is the only significant non-means-tested form of direct income assistance for working-age adults.5 The Federal-State Unemployment Insurance Program is financed by employers (and, in three states, by employees), with employers’ contributions experience-rated to reflect the number of prior claims from their establishments. Although most workers are covered by UI, benefit eligibility is based on prior earning requirements and nonmonetary requirements that exclude many from claiming benefits and reduce the effectiveness of UI as an antipoverty program. Only about 37 percent of unemployed adults received Unemployment Insurance through federally mandated, state-run programs as of 2007, and the share of the unemployed receiving benefits has declined in recent years (Stone, Greenstein, and Coven 2007). UI is designed to provide job search and cash assistance during temporary periods of joblessness, usually up to a maximum of twenty-six weeks. Not all workers qualify for the maximum coverage period and even those covered weeks are not enough for many workers; as many as 40 percent of UI recipients exhaust their benefits (Vroman 2003). Stringent eligibility criteria create particularly great barriers for lowearning parents, who are less likely than other workers to receive benefits during periods of joblessness because they fail to accumulate enough earnings, have an intermittent work history, have been part-time workers or are seeking part-time employment, or have left employment voluntarily for personal or family reasons that do not qualify as “good cause” to receive UI payments. Examining data for the 2001-to-2003 economic recession, Gregory Acs, Harry J. Holzer, and Austin Nichols (2005) estimate that only 12 to 14 percent of poor households received UI

270

/

Beyond the Safety Net

during the economic downturn and that the program played a relatively small role in reducing poverty among families with children or slowing the increase in their poverty during that recession. The second component of United States income benefits is rooted in the tax system. Unlike nearly all of the countries in Europe, the United States does not provide a direct child or family allowance to parents. Parents benefit instead from a number of tax benefits, some targeted on families with children and others not. As antipoverty policy, tax benefits have limited potential because nearly all are nonrefundable for those with low or no tax liabilities. The standard income tax deduction for resident children, for example, does not benefit low-income families who do not have tax liabilities. The majority of low-earning parents are also unable to benefit from specialized tax deductions, say, for home mortgage interest, or from tax subsidies for employer-provided benefits: 65 percent of workers in the lowest wage quintile do not have employer-provided health insurance and 80 percent are not enrolled in an employer-sponsored pension plan (Gould 2006; U.S. Department of Labor 2008). Low-earning parents in the United States benefit most from two targeted tax credits. Some qualify for the partially refundable Child Tax Credit (CTC), which provides custodial parents with a credit of as much as $1,000 for each child under the age of seventeen. The CTC was designed to benefit middle and low earners, phasing out for higher-income families, and is not fully refundable, thereby excluding many of the lowest earners. In 2005, less than 5 percent of all CTC benefits were estimated to have gone to tax-filing units with incomes below $20,000 and benefits for these filers averaged less than $1,000 (Burman and Wheaton 2005). The single largest program of cash assistance for low-income families in the United States is the refundable federal Earned Income Tax Credit (EITC). As of 2006, twenty states also provided earned income tax credits that provide between 15 and 30 percent of the federal credit (Holt 2006). Earned Income Tax Credits reflect both the emphasis on employment and the targeting of assistance in the United States: until the phase-out range, the credit increases with earnings, and unlike the CTC, most benefits—an estimated 70 percent in 2003—go to filers with incomes of $20,000 or less. The third, the most visible and arguably most controversial, component of the income benefit system for families in the United States is cash assistance through the Temporary Assistance for Needy Families (TANF) program. Participation of poor families in means-tested cash assistance programs declined dramatically, and spending was shifted from cash benefits to services, following policy changes of the mid1990s that required states to impose lifetime time limits on receipt of any cash assistance, stringent work requirements, and financial sanctions for nonparticipation in work activities. Between 1994 and 2002, the share of all pre-tax and pre-transfer poor families with children who received cash assistance declined from 77 to less than 40 percent and the value of benefits dropped by about 22 percent (Meyers et al. 2008). THE NORDIC APPROACH The Nordic approach to income benefits for families combines an emphasis on full employment for mothers and fathers with a

/

271

Making the Work-Based Safety Net Work Better

universalistic, inclusive approach to income support that minimizes complexities and discontinuities associated with changes in employment, earnings, and family structure. As in the United States, income benefits are provided through employment-conditioned unemployment compensation, targeted child and family benefits, and assistance predicated on income or single parenthood. Unlike the United States, these programs are designed to be broadly inclusive and many are available for all families with children. Although their programs differ in their details, the Nordic countries are among the most comprehensive and inclusive among the OECD countries in the design of benefits for jobless workers, combining contributory Unemployment Insurance (UI) programs, which replace 80 to 90 percent of wages for covered workers for as long as two to four years (in Norway and Denmark, respectively), with Unemployment Assistance (UA) programs that provide means-tested assistance for job seekers who fail to qualify for regular UI or exhaust their benefits.6 The Nordic programs define employment broadly for the purpose of qualifying for UI to include paid employment, self-employment, and some periods of employment preparation or paid leave for family reasons. The Nordic programs also base eligibility on prior working time rather than earnings, thereby narrowing the access gap between lower- and higher-earning workers. Unlike the UI programs in most U.S. states, which condition eligibility on minimum earnings, these systems grant eligibility to those who meet minimum requirements for hours, days, or weeks worked within the base (or qualifying) period prior to application for benefits. In some countries the base or qualifying period for meeting minimum work requirements is quite long, extending coverage to job seekers with intermittent work histories. In Denmark, for example, applicants must have twenty-six weeks of employment in the prior three years; in Finland, employees must have worked twenty-six weeks in the prior two years, and self-employed persons must have at least twenty-four months of “entrepreneurship” during the prior forty-eight months. Using working time rather than earnings extends coverage to a larger share of the low-wage workforce and eliminates inequities in eligibility arising from earnings differentials: a low-wage worker who meets the minimum time requirements has the same eligibility as a high-wage individual meeting the same minimum working time requirement. UI also reaches more workers in the Nordic countries because penalties for voluntary job terminations are less severe and coverage is extended to part-time workers. The United States imposes heavy penalties on workers who leave their jobs without “good cause”; in most states they are disqualified for the duration of the unemployment period. This imposes a particularly heavy penalty on women and on parents because in most states family issues, such as the loss of child care, do not qualify as “good cause.” Although the Nordic countries also impose penalties for voluntary quits, they are considerably less severe—benefits are suspended for ten to sixty days for unemployment resulting from a voluntary leaving and reduced by 25 to 50 percent for refusal to accept a suitable job. The UI systems in the Nordic countries are also more inclusive of part-time workers than are the systems in the U.S. states. Coverage for part-time workers is particularly important

272

/

Beyond the Safety Net

for the economic security of families with children because part-time workers are disproportionately women in all of the industrialized countries—typically women who are combining employment with the care of young children. In the United States, individuals who work less than full-time are typically unable to claim UI when they lose their jobs, because they cannot meet the minimum earnings requirements; jobless individuals who seek part-time work are also disqualified from UI in most states. In the Nordic systems, the use of days or weeks of prior work to establish eligibility for UI has the effect of extending eligibility to parttime workers because “part-time” days and weeks are counted. For example, in Denmark, part-time workers are eligible for benefits if they have worked at least fifteen hours per week for seventeen weeks during the three years before seeking benefits. In Sweden, gainful employment is required for at least eighty days and workdays of at least three hours are counted. The second component of income benefits, child and family allowances, are provided directly in Nordic countries and are among the most generous in the OECD countries. Unlike the targeted, earnings-based EITC in the United States, benefits are universal with no employment or income eligibility tests. Child and family allowances are a substantial addition to family incomes and are a particularly important component of the income package for low-income families. In Sweden, Finland, and Norway, benefits for low-income families average between 9 and 12 percent of median equivalent income per child; in Denmark, average benefits are somewhat lower at 10 percent (Rainwater and Smeeding 2003). Universal benefits are supplemented for the most economically vulnerable families. In Denmark, for example, the basic family allowance is supplemented for families with a single earner or headed by a single parent. In Sweden, universal child allowances are supplemented for large families and for single parents. The Nordic countries provide additional assistance to single-parent families through the provision of advance maintenance payments when noncustodial parents do not pay child support. Rainwater and Smeeding (2003) estimate that 82 and 89 percent of single parents in Norway and Sweden, respectively, receive child support from either the absent parent or the government, and these benefits average 8 to 10 percent of median income. When combined with child and family allowances, child support in these countries provides income for single-mother families equivalent to about one-half of the poverty line (measured as 50 percent of median household income). Finally, universal family allowances and child support are supplemented by means-tested transfers for the poorest families. Two-thirds to three-quarters of all poor single-parent families receive some means-tested income assistance in the Nordic countries, and a somewhat smaller share of poor two-parent families receive public assistance. In combination with other forms of assistance, these benefits bring the poorest families close to one-half the median income in these countries, a common measure of poverty. Whiteford and Adema (2007) compare post-tax and post-transfer incomes of the poorest families in OECD countries— those receiving means-tested assistance—to median disposable incomes in their countries. On average, single-parent and two-parent recipient families have in-

/

273

Making the Work-Based Safety Net Work Better

comes that are about 33 percent of their countries’ median incomes across the OECD countries. In the United States, their incomes are very low, about 20 percent of the median, whereas in Sweden and Denmark they are over 40 percent.

Time for Caring Policies that provide job-protected, paid family leaves are a critical form of support for helping parents reconcile earning and caring roles.7 Family leaves promote economic security directly by assuring income during periods—such as the months following the birth or adoption of a child and days of child illness—when parents cannot be at the worksite. They promote economic welfare over the longer term by encouraging parents to remain attached to employment and employers to retain jobs for parents during temporary absences for family reasons. Family leaves raise questions about financing that have been particularly contested in the United States. They also raise serious questions about gender equality if they encourage mothers, but not fathers, to loosen employment ties when their children are young. THE UNITED STATES APPROACH Perhaps more than any other family-policy area, the U.S. approach to protecting parental time for caring reflects a minimalist, private, and market-oriented approach to social benefits that stratifies assistances and excludes low-earning families. The United States is an exceptional case, in cross-national terms, due to the absence of any national policy granting job-guaranteed, paid leave at the time of childbirth or adoption.8 The national Family and Medical Leave Act (FMLA) of 1993 requires large employers to extend job-protected but unpaid leaves following childbirth and adoption and for serious family illnesses. Although there is no national program of social-insurance–based maternity, paternity, or parenting leave, three states (California, New Jersey, and Washington) have recently passed laws granting brief periods of paid parenting, or “bonding,” leave following birth and adoption, with modest benefits paid through employment taxes. Five states provide some paid maternity leave through state Temporary Disability Insurance (TDI) programs that are legally required, under the 1978 Pregnancy Discrimination Act, to treat childbirth as a disability. Mothers in the remaining states, and fathers nearly everywhere, have to rely on paid leave granted voluntarily by their employers. And unlike other employer-provided benefits, such as health care and retirement pensions, these private employment benefits are not subsidized through the federal tax system. In the absence of job protections and wage replacements, most low-earning parents have no realistic options for adjusting their employment to care for infants or ill family members. The share of workers with unpaid family leave coverage increased in the 1990s with the passage of the FMLA (Waldfogel 1999). Even with this expansion, one-half of part-time workers and full-time workers in small firms remained uncovered and the share with paid coverage has remained stubbornly low. Only about 25 percent of employers continue workers’ full pay during FMLA-covered leaves and 25 percent offer partial pay or pay under certain cir274

/

Beyond the Safety Net

cumstances; about 50 percent of employers provide no pay during parental leaves (U.S. Department of Labor 2000). The absence of wage replacement during FMLA-guaranteed leaves means that many families have no source of replacement income during leave periods and others cannot afford to take the leave to which they are entitled. Nearly 80 percent of employees who do not take FMLA leave when needed report that the reason is that they “could not afford to take leave” (U.S. Department of Labor 2000). Others cut back on their leave period; among parents who take FMLA-guaranteed leaves to care for newborns, over two-thirds take fewer than half of their allowable days. Lacking wage replacement, the FMLA benefits some parents with middle and high incomes but is of little help to low earners. Fewer than half (46 percent) of poor working parents have any paid maternity leave (Phillips 2004) and about 10 percent of FMLA users who do not receive their full pay go on public assistance during their leaves (U.S. Department of Labor 2000). Although the FMLA is working poorly as antipoverty policy, in cross-national terms it is somewhat progressive with regard to gender equality.9 The FMLA grants mothers and fathers equal and separate rights to take time away from the job following childbirth or adoption, or to attend to a family illness. This policy design helps to equalize the incentives for mothers and fathers to take time off and to share caregiving work. However, the lack of wage replacement deters many men from taking leave, especially in two-parent families in which they are the higher earner. THE NORDIC APPROACH The Nordic countries—Norway, Sweden, Denmark, and to a lesser extent Finland—provide a model for a very different approach to paid family leave that emphasizes principles of universalism, protection of parental caregiving time, early child development, and gender equality. Each of these countries has institutionalized a public social insurance program that is available to low-earning parents along with their more-advantaged counterparts, reducing the risk that a pregnancy or family illness will push a working family into poverty. The systems vary across these countries in a number of ways, but they share several common features. In each of these countries national maternity-leave policies grant nearly all employed mothers several weeks or months of job security and wage replacement around the time of childbirth or adoption. Family leave policies in the Nordic countries offer mothers the equivalent of about thirty to forty-two weeks of leave with full pay, typically up to an earnings cap. These countries achieve high levels of provision through various mechanisms. In Norway and Sweden, maternity and parental leave are blended into a single program that grants couples an allocation of about a year to be shared between them; wage replacement is high for the whole period, at 80 to 100 percent. Finland and Denmark offer eighteen weeks of maternity benefits (at about two-thirds pay, on average), followed by separate parental-leave options that couples may allocate to the mother if they choose. In Denmark, collective agreements compel many employers to “top up” public benefits so that, in practice, most workers receive their full pay. To contain costs, benefits are limited or capped for the highest-earning parents. /

275

Making the Work-Based Safety Net Work Better

Finland, for example, reduces the replacement rate stringently as earnings rise. Norway and Sweden place caps on covered earnings, but the caps are set high— respectively, at about 1.9 to 2.2 times average earnings among mothers of working age, including both part-time and full-time workers. Earnings caps result in a progressive benefit structure and restrain program costs; when caps are set high, most mothers and their families are protected from substantial losses in economic security during leave periods. Most employed parents in the Nordic countries also have the right to take relatively long periods of leave—usually a full year or more—and, through social insurance funds, they receive about two-thirds or more of their wages during most or all of their leave periods (again, subject to caps for high earners).10 Parental-leave policies in the Nordic countries afford parents substantial flexibility: Denmark and Sweden allow parents to take their allotted paid leaves in increments until the child is eight years old; Norway and Sweden allow parents to combine pro-rated leaves with part-time employment; and Finland and Norway permit parents to use a portion of their leave benefits to purchase private child care instead.11 Gender differentials in the use of leave have been a persistent problem even in countries that extend benefits to fathers. Although none of the Nordic countries have achieved gender equality in leave usage, several are taking steps to increase fathers’ use of leave benefits. Two strategies appear particularly promising. First, high wage-replacement rates provide incentives for fathers, who are usually the higher earners in couples, to take their allotted leave benefits and protect families from a loss of income when they do so. Second, nontransferable individual rights and earmarking some portion of a family-based entitlement for each parent separately create “use or lose” provisions that increase the incentives for fathers to make use of leave; leave time that is not taken by the father is lost to the family. In each of these countries, these leave schemes are financed through social insurance mechanisms. Benefits are funded by employee and employer contributions, often supplemented by general tax revenues. A key lesson is that none of them relies on individual families to finance leave, nor on mandating employers to provide wage replacement for their own employees. Where social insurance financing does depend heavily on firms’ contributions, those contributions are independent of employees’ usage rates. Social insurance financing distributes the burden for employees and employers. The costs of caregiving are shared across employees’ working years, among parents and nonparents, between leave-takers and non-leave-takers, and across enterprises as well. These financing mechanisms, especially where supplemented by substantial contributions from general tax revenues, reduce the risk for individual families and individual employers. They reduce employers’ resistance and lessen incentives to discriminate against potential leave-takers.

Substitute Care for Children Alternatives to full-time parental care that provide high-quality and developmentally appropriate care are an important component of the social benefits package

276

/

Beyond the Safety Net

for all families with children and an essential benefit for low earners. Parents cannot work for pay unless they have care for their children during working hours and mothers, in particular, will reduce their time at work if care does not meet their standards for quality. Policies that subsidize the costs of high-quality care are particularly crucial for the economic security and work incentives of lowearning parents because they free them for employment and reduce the loss of earnings to child care costs. Public subsidies that raise caregivers’ wages also prevent the creation of a low-wage workforce of child caregivers. Public involvement in the care of children also raises trade-offs between issues of quality and cost, and sometimes complex questions about public and private responsibility for children and about the diversity of family values and childrearing beliefs. THE UNITED STATES APPROACH The United States historically defined the care of children before school age as a private responsibility. While many of the countries of Europe were developing public child-care and preschool systems, the United States developed an extensive private market system of child-care centers, family child-care homes, and sitters. In recent years, public involvement and public expenditures for early-childhood education and care (ECEC) have grown substantially in the United States, in response to both increasing rates of maternal employment and increasing concern for early learning and school readiness. As of the early 2000s federal and state governments spent over $9 billion on meanstested child-care subsidies and as much as $7 billion on targeted preschool, prekindergarten, and Head Start programs; they also “spent” over $3.5 billion on taxbased child-care benefits through the (nonrefundable) Dependent and Child Care Tax Credit and tax-exempt flexible spending accounts arranged with employers (Blau and Currie 2006; Magnuson, Meyers, and Waldfogel 2007). Recent expansion of state-funded pre-kindergarten programs reflects concerns about children’s school readiness and have been promoted, although not yet adopted, as an extension of free universal education. In contrast, support for public investments in other forms of subsidized care for low-income families—from the care of infants to child-care centers and before- and after-school programs— has been driven largely by the employment-forcing approach of United States social welfare policies. As a result, the benefits provided to the lowest earners are often insufficient to cover the costs of the best-quality care and are strictly limited to care during parents’ working hours. Restrictions on the use of subsidies are particularly problematic for low-earning parents because they are more likely than other workers to have intermittent employment and variable, nonstandard, and unpredictable working schedules. Although government expenditures have grown in recent years, particularly for the poorest families, assistance remains limited, fragmented, and ultimately very regressive in the distribution of all child-care costs. Targeted services are available for only a fraction of income-eligible families—an estimated 15 to 25 percent of income-eligible families receive child-care subsidies and about 17 percent of four-year-old children are enrolled in pre-kindergarten programs (U.S. Department of Health and Human Services 2000; Barnett et al. 2005). As a result,

/

277

Making the Work-Based Safety Net Work Better

out-of-pocket costs remain disproportionately high for the lowest-income families who use child care. Although low-income families pay less on average for child care than do more affluent families, largely because they rely more heavily on “free” care donated by family and friends, child-care costs consume a larger share of their income: among employed families who pay for care, low-earning families pay an average of 16 percent of earnings for child care, in contrast to the 6 percent paid by higher-earning families (Giannarelli and Barsimantov 2000; Giannarelli, Adelman, and Schmidt 2003). THE NORDIC APPROACH Integrated “EduCare” systems in the Nordic countries are the most comprehensive public child-care systems in the OECD. There is very little private-market care, and in Denmark, Finland, and Sweden, children have a right to a place in public child-care and preschool programs. This care is an essential component of the social democratic commitment to full employment, gender equality, and reduction of income and social inequalities.12 Rather than a trade-off with parental care, high-quality public child care is seen as an essential contribution to the health and education of young children. In the Nordic countries public systems under the authority of national social welfare or educational authorities serve children from the end of paid family leave periods until the start of primary school. Younger children are cared for in centers or supervised child-minder arrangements; older children may spend all or part of their day in preschool programs. These systems are most notable for extending a nearly universal entitlement for care (with a modest parental copayment) during the years before the start of primary school; parents have a right to a place in a public child-care setting and the regular use of fully private care is rare. The Nordic countries finance ECEC through the direct provision of public care, cost-sharing with parents through co-payments, and the use of alternative financing mechanisms such as demand-side subsidies and tax benefits. National tax revenues cover about one-quarter to one-third of the costs of ECEC and municipal governments contribute about one-half. Parent fees cover a capped share of the costs. The parental share varies across countries and with the type of care, averaging about 20 percent. Fees for individual families are calculated on a sliding fee scale and often are waived altogether for low-income families. Balancing affordability with service quality is a key challenge in the provision of early-childhood education and care. Quality of care is important for parents, whose ability to engage in market work depends on their trust in the care that their children are receiving while they are the workplace. And quality of care is essential for the healthy development of children. The Nordic countries encourage high-quality care through a combination of direct oversight and provision of services and through regulations for the qualifications and compensation of ECEC workers. All but Finland require bachelor-level university degrees of both child-care workers and preschool teachers; Finland requires a university degree for preschool teachers and a three-year vocational or polytechnic degree for child-

278

/

Beyond the Safety Net

care workers. In Sweden, 98 percent of child-care workers have specialized certification or university degrees.

Policy Developments in the United Kingdom In this chapter we have compared the U.S. social benefits package for families with the comprehensive and universalistic approach of the Nordic countries. This comparison is useful for illustrating the institutional design and feasibility of a system that is substantially different from the current U.S. approach. Cross-national comparisons can also be useful when “most similar” cases are compared in order to identify possibilities for more incremental change in policy institutions. The United Kingdom is typically grouped with the United States, and with other English-speaking countries, in cross-national comparisons of social benefits and poverty. Both countries have historically provided limited social benefits; both have historically had rates of child and family poverty that exceeded those in other industrialized countries by a wide margin. The two countries have differed in patterns of maternal employment in recent years, however, with a much sharper rise in employment, particularly full-time employment, among women in the United States. In the final years of the twentieth century, the United Kingdom began to chart a policy course different from that of the United States. In 1999, Prime Minister Tony Blair announced an ambitious goal to end child poverty in the United Kingdom within a generation. This promise stood in sharp contrast to U.S. policies of the period, which focused on ending welfare rather than poverty. Depending on how poverty is measured, the United Kingdom was much more successful than the United States in reducing child poverty in the following years. The U.S. child poverty rate stood at just over 17 percent as of 1999, using the federal poverty threshold, which corresponds to about 35 percent of median income; by 2006 the rate had inched up to nearly 18 percent. During this same period, child poverty declined in the United Kingdom by 18 percent when poverty is measured using the relative measure that is standard in the United Kingdom: income at or below 60 percent of the median. Applying an absolute measure to the U.K. data that sets a threshold at 60 percent of median income as of 1997 with adjustments for inflation, the decline in poverty is a more substantial drop in poverty from 22 to less that 13 percent by 2005 and 2006 (Waldfogel 2007). The United Kingdom achieved this reduction in child poverty through a combination of direct transfers, tax credits, expansions to paid family leave, and child care that is moving the country closer to the Nordic model. The U.K. reforms included three main elements: policies to promote paid work and make work pay; other policies to raise incomes for low-income families with children; and direct investments in children (Waldfogel 2007). In terms of the three policy areas discussed in this chapter, the most significant reforms to the U.K. social benefits package have increased income benefits (Wald-

/

279

Making the Work-Based Safety Net Work Better

fogel 2007; Minoff 2006). Policies to increase work effort have included New Deal–type programs that provide employment assistance for young adults, for lone parents, and for disabled adults. Provisions for increasing the value of employment included the introduction of a minimum wage and the Working Tax Credit, an employment-based tax credit similar to the EITC in the United States. The basic U.K. Working Tax Credit is now available to all low-income workers and includes additional elements for disability, for care of children, for lone parents, and for adults working thirty hours or more per week. An additional Childcare Tax Credit is designed to offset a portion of parents’ out-of-pocket expenditures for couples or lone parents who work at least sixteen hours per week. In addition to employment-based supports for low earners, the United Kingdom provides a floor under the income of all families with children through universal child allowances and tax benefits. And unlike in the United States, both universal and means-tested assistance were expanded as part of recent policy reforms. The United Kingdom expanded transfers through the existing universal Child Benefit and increased benefits for families with young children through the Child Tax Credit, available to nearly all families, and through means-tested Income Support benefits for those with little or no earned income. Relative to the United States, the United Kingdom has substantially more generous income benefits for both employed and nonemployed low-income parents. Provisions in the other areas we consider in this chapter are also closer to those of other European countries, although less comprehensive than those of the Nordic countries. The United Kingdom has had paid maternity leave in place for many years and recently expanded Statutory Maternity Pay for women with substantial employment histories, and more broadly available Maternity Allowances from twelve to twenty-six weeks. Both programs provide a period of wage replacement followed by additional weeks with a flat benefit. As of 2003, fathers became eligible for two weeks of paid paternity leave under the same wage replacement rules as Statutory Maternity Pay. Each parent is also eligible for up to thirteen weeks of unpaid, job-protected leave. For many years the United Kingdom trailed other European countries in public child-care provisions. The renewed commitment to reduce child poverty led to a substantial increase in child-care assistance. The majority of this assistance is provided through publicly subsidized, part-day preschool services, which are now available for all four-year-old children and are expanding to serve three-year-olds as well. Part-day services do not provide sufficient care for full-time employment, however. The U.K. has also expanded means-tested child-care subsidies for the lowest-income families participating in employment support programs.

SUMMARY AND RECOMMENDATIONS In comparison to the other industrialized countries of the OECD, and particularly to those of northern Europe, the United States has higher employment rates, longer average working hours, and a smaller proportion of poor children living

280

/

Beyond the Safety Net

with families with no working adults. In spite of this work effort, more families with children are poor in the United States than in nearly all of the other rich OECD countries. Why is the United States trailing so badly in protecting the economic security of families? One answer can be found in the share of adult workers earning low wages, which is higher in the United States than in many of the rich industrialized countries. Another answer is that the United States also does less to ensure the economic security of families in which parents earn low wages. The United States does provide a relatively generous and sustainable social benefits package to supplement market earnings. For working-age adults, these benefits are provided through tax- and employment-based benefits, including child deductions and credits on federal and state income taxes, additional deductions for home mortgage interest and out-of-pocket child-care expenditures, taxsubsidized private health insurance, and other employment benefits such as jobprotected family leave and paid sick leave. These benefits are usually stable over time and the value of many increases with earnings. Most of these benefits are available only to those with steady employment and moderate-to-high incomes, however. The social benefits package for families with low-earning parents could not be more different. Families in which parents have intermittent, part-time, or low-wage work are less likely than those with higher earners to qualify for either Unemployment Insurance or Old Age, Survivors, and Disability Insurance (OASDI) during periods of joblessness. They are exempted from most income taxes by progressive federal and state tax schedules, but these same policies exclude them from nonrefundable, tax-based benefits. They are unlikely to have employer-provided health care or other benefits. For most families headed by low earners, the social benefits package is provided through targeted, means-tested welfare programs. These benefits are designed to be temporary and decline in value as earnings rise. For most families, income-tested benefits disappear long before these families are able to secure sustainable assistance through employment-based benefits or nonrefundable tax benefits. In the wake of the 1990s welfare reforms, the challenge for United States policymakers is not motivating adults into the workforce or reducing welfare receipt. The challenge is ensuring the economic security of millions of families in which parents work but remain poor or very low-income on the basis of income alone. One step in meeting this challenge should be the design of a social benefits package for low-earning adults that does not function as a traditional welfare or safetynet program but instead provides the same accessible and sustainable support over time that is currently provided through tax- and employer-based benefits for moderate and high earners. A number of observers have suggested the components of a social benefits package or an employment support system for working families. Rather than cataloguing specific benefits, in this chapter we have considered the structure or architecture of a subset of social benefits and employment supports—as they exist now in the United States and as they have been developed in several of the countries of northern Europe, with an emphasis on the integrated and comprehensive systems that have been developed in the Nordic countries and may be emerging

/

281

Making the Work-Based Safety Net Work Better

in the United Kingdom. From this exercise we distill the following lessons for the institutional design of a social benefits package to support the long-term economic security of working but poor families in the United States • Integrate income security, antipoverty, employment support, and work-family reconciliation measures into a coherent package for working families. The United States’ approach to social and employment policy is exceptionally fragmented across programs that are narrowly targeted to the poor, or to single mothers, or to middleclass families. This fragmentation leaves important gaps in benefits and coverage and increases administrative costs for government and employers. It also creates oftentimes unintended interactions among public, private, and tax-based benefits that penalize low-earning families as they increase their work effort. In their recently published review of work and family policy OECD researchers conclude that policy systems “that provide a continuum of support to families—support for parents at home when the child is very young, leading on to a child care place, preschool, school and out-of-school-hours care activities—performs best in helping parents reconcile work and family life” (OECD 2007a, 18). In combination with other universal benefits, most especially health care and housing assistance, these social benefits packages assure the economic security of all families and dramatically reduce market-generated poverty. • Provide benefits through integrated systems that do not institutionally separate programs for lower- and higher-earning individuals. In each of these policy areas, and in others that we do not review here, our European counterparts provide broad coverage with eligibility based on categorical criteria such as the presence of children, or on joblessness rather than on prior earnings, income, or assets. Broad, non-means-tested coverage decreases the stigma and increases participation in social benefit and employment support programs. And even more important for working but poor families, benefits that are not limited by income or assets are sustainable as number of hours worked, wages, and earnings change; in other words, parents and their families do not gain or lose essential coverage and benefits nor do they have to change programs with changes in employment status or market income. Broadly inclusive family allowances, fully refundable tax credits, advance child support payments, subsidized child care, and funded family leaves provide a solid base that, in combination with earnings, assure a minimum and continuous level of economic security for children and their families. • Adjust universal benefits by income to provide extra assistance to the poorest and to cap benefits for the most affluent. Although they do not impose income tests, the model systems of the Nordic countries that we describe in this chapter do adjust some benefits to family income—for example, they exempt low-income families from co-payments for early-childhood education and care, supplement child and family allowances for lone parents, and “top up” leave benefits to reach a minimum threshold. To contain costs, benefits such as family leave are often capped or may be subject to taxation for the highest-income recipients. Taxation of benefits for the highest earners, rather than means testing of benefits for the lowest earners, can achieve income targeting and cost containment goals without penalizing low-earning parents as they increase their earnings and assets. Taxing back bene-

282

/

Beyond the Safety Net

fits at the top is also less administratively burdensome and costly than targeting benefits at the bottom, which typically requires detailed eligibility rules and bureaucratically cumbersome intake procedures. • Finance benefits through public tax and transfer or social insurance systems rather than through employer-based arrangements. Employers in the northern and western European countries, like U.S. employers, provide some portion of the total social and health benefits received by workers. However, these private arrangements typically augment, or “top up,” core social insurance or other public benefits that are available to all eligible individuals. Private benefits in other OECD countries are also negotiated within the context of country- and EU-level oversight of employment contracts and conditions. In the United States, where government involvement in the terms of employment is notably weak, reliance on employerprovided benefits has created an exceptionally regressive tier of private benefits for moderate and high earners that includes health care, pensions, family leave, sick leave, and other benefits. These benefits are negotiated through private employment contracts but a substantial share of their costs is subsidized through their exclusion from employer and employee taxes. In our European comparison countries, particularly the Nordic countries, public financing and regulation of benefits through social insurance and other tax systems has the effect of distributing benefits more broadly and costs more progressively with earnings. Working parents in these countries do not have to move from welfare-based benefits to taxbased and private benefits as their earnings rise, with the concomitant risk that they will fall into a gap between these systems and lose benefits altogether. • Finance benefits through national social insurance systems or general revenues that pool and redistribute resources across households and across the life cycle. The public sector has the capacity to redistribute and equalize benefits across families that have higher and lower market earnings and can command more and less generous nonwage compensation from employers. But only the federal government has the “deep pocket” necessary to provide adequate and sustainable funding. Centralized national financing also reduces inequalities across states and regions with more and fewer resources. The United States has well-developed, centralized, and efficient systems for the elderly and some disabled through the OASDI social insurance programs. In sharp contrast, assistance for families remains decentralized and fragmented. The United States has a long history of state-level policy diversity and innovation. The most significant of these local innovations, however, have eventually been centralized through national programs that have broad coverage and more adequate financing. While encouraging innovation, it is critical to move toward national financing and oversight of the key components of the social benefits package for families in the United States.

FINAL THOUGHTS The United States social and employment benefit system is not functioning to help working families achieve and maintain economic security. In this chapter we

/

283

Making the Work-Based Safety Net Work Better

have argued that this is a matter of how the United States spends and structures these benefit packages for families, not just of how much it spends. Here we take up two questions about the potential costs and consequences of changing these structures. The first question is one of unintended consequences. Would more generous benefits—for example, through more comprehensive Unemployment Insurance coverage and paid family leave programs—have the perverse effect of lengthening the amount of time spent outside the workforce and thus increasing family poverty? There is good reason to question the impact of more inclusive and generous benefit programs on work effort and earnings. Economic theory and empirical research suggest that making income benefits more available and more remunerative may lengthen spells spent outside the workforce by creating a viable alternative to employment. The increase in the period of joblessness may be modest, however. By one recent estimate, for example, unemployed workers in the United States who collected benefits stayed unemployed for only about two and one-half weeks longer than those who did not (Needles and Nicholson 1999). A more fundamental question about paid unemployment, parenting, or other temporary absences from the workplace is whether these periods lead to better or worse outcomes in the longer term. Research on Unemployment Insurance suggests that the relatively modest increases in the length of unemployment spells due to more generous benefits are offset by the benefits of a longer period of job search. A more inclusive and generous unemployment benefit system could improve economic outcomes for low-earning families by giving parents time to upgrade skills and to find jobs with higher pay and needed benefits. Recent studies have found that unemployed workers who collect benefits receive higher pay when they return to work than those who do not, and also that receipt of unemployment benefits raises the likelihood that the new job will have employer-sponsored health insurance (see Emsellem 2006 for a discussion of this research).13 In the case of paid family leave, research does suggest that public policies that grant very long periods of labor-market withdrawal for child care can reduce women’s labor-force participation and lifetime earnings substantially (OECD 2001). Shortterm paid leaves have the opposite effect, however, increasing the probability that women return to the workplace by the end of the postpartum year and the probability that they will be employed in later years (OECD 2001). More generous and inclusive social and employment benefits for families, similar in structure to those in many of the countries of northern Europe, could increase economic security and move child poverty rates in the United States closer to the levels observed in these countries. The feasibility of adopting such a system raises a second question. How much would it cost to move the United States to a publicly financed system of social and employment benefits that supplements low earnings, provides parents with time to care for young children, and subsidizes the costs of high-quality early-childhood education and care? Public financing for the first of these forms of support, supplementing low earnings to assure a minimum level of income security, could be accomplished

284

/

Beyond the Safety Net TABLE 8.1

/ Estimated Direct Costs: Take-Up of Paid Family Leave and EarlyChildhood Education and Care Programs

Type of Program Paid family leave High take-up (100 percent) Moderate take-up (approximately 50 percent) Early-childhood education and care (ECEC) High take-up (100 percent of children) Moderate take-up (50 percent of children under three) Total High family leave take-up and high ECEC take-up High family leave take-up and moderate ECEC take-up Moderate family leave take-up and high ECEC take-up Moderate family leave take-up and moderate ECEC take-up

Cost in U.S. Dollars (billions)

Percentage of U.S. GDP

45.0 22.5

0.43 0.22

111.1

1.07

84.4

0.81

156

1.50

129

1.24

134

1.28

107

1.03

Source: Authors’ calculations.

through reforms to the Unemployment Insurance system and consolidation of existing means-tested welfare transfers and various specialized tax deductions and credits currently provided to parents and to low earners. Whether a more inclusive system for supplementing the income of low-earning parents imposes new public costs will depend on benefit levels and the extent to which government regulations and worker demands set a floor under wages. Financing for the second and third elements of this proposal—paid parenting leaves and high-quality early-childhood education and care—would involve new spending by the United States government. How much? As a thought experiment we estimate the direct costs of creating a comprehensive and generous system of universal paid family leave and early-childhood education and care services for United States families (see table 8.1).14 We estimate that the total direct cost of paid family leave would be approximately $22.5 billion to $45 billion per year in new social insurance spending, depending on the level of take-up. These costs are based on the provision of a six-month benefit for both mothers and fathers, payable at 100 percent wage replacement (with an earnings cap of about $69,000 per year in 2004), high take-up rates, and no minimum enterprise size. The lower-bound estimate assumes that leave-takers claim an average of 50 percent of the days to which they are entitled; the upper bound assumes that they take up the entire period to which they are entitled.

/

285

Making the Work-Based Safety Net Work Better

Universal early-childhood education and care—for about twenty-six hours a week on average for children under three, full-time for three- and four-year-olds, and half-time for five-year-olds (the majority of whom are now enrolled in public kindergarten part-day)—would cost an estimated $111 billion if government assumed 82 percent of costs and take-up was 100 percent among families. If family take-up for one- and two-year-olds was closer to the 50 percent of Swedish families, the total public cost would be about $84 billion. Given current federal spending of about $16 billion on subsidies, tax benefits, and Head Start, the programs would require an estimated $95 (high take-up) to $68 (low take-up) billion in new spending for the federal government—and would free up additional funds that states are currently devoting to child care and early education. Are these direct costs a lot or a little to spend on the well-being of families and children? This level of spending would be comparable to what some of our European counterparts now invest in these programs. The paid leave, child care, and early education benefits combined represent about 1.0 to 1.5 percent of the current United States GDP for moderate take-up and high take-up, respectively. In comparison, Sweden spends about 2.5 percent of its GDP on the combination of family leave and ECEC. The United States currently spends less than one-tenth of this amount, approximately 0.15 percent of its GDP on publicly financed child care and a negligible amount on publicly financed paid leave.15 As the richest country in the world, with one of the highest child poverty rates among developed countries, the United States clearly seems positioned to do more than it is currently doing to support the economic security of its families with children.

NOTES 1. Workplace regulations that increase flexibility and accommodations for parents’ needs are another important component of work-family reconciliation policies but are beyond the scope of this chapter. Several recent studies find that low-income working parents typically have substantially less access to measures that help parents reconcile work and family responsibilities than do their higher-earning counterparts. A growing literature finds, in particular, that lower-skilled, lower-waged workers are the least likely to have options for shifting their work schedules, in either the short term or the long term. Their lack of access to scheduling flexibility makes it difficult to respond to family emergencies and unexpected needs; the inevitable conflicts that arise make these workers especially vulnerable to job loss (see, for example, Dodson and Bravo 2003; Williams 2006; Levin-Epstein 2006). 2. Tax subsidizes for other child- and family-related benefits are both more recent and more limited in scope. A portion of private child-care expenditures can be credited against individual earned income or set aside as pre-tax income by employers, but other workplace benefits, such as maternity and parental leave, confer no tax benefits for either employers or employees. 3. The value of the package is estimated using survey responses for the receipt of cash and in-kind benefits and the use of specialized tax deductions, respondent’s estimate

286

/

Beyond the Safety Net

4. 5.

6. 7.

8.

9. 10.

11.

12.

13.

14.

of the value of cash benefits, and imputed values for in-kind benefits based on government cost-accounting methods. To compare the value of public and publicly subsidized private health insurance the analysts include both the estimated tax benefits for employer-provided health insurance and the market value of that coverage. As noted earlier, this discussion sets aside the critical area of policies that increase earnings, which is discussed elsewhere in this volume (see chapter 2 by David Neumark). An estimated 5 million or more children also benefit from social insurance (through the OASDI) either as beneficiaries or as members of households that receive social security or disability insurance income. As Nancy Cauthen (2005) observes, “Social Security is a family insurance program” and more children benefit from OASDI than from TANF or SSI. For a more detailed review of these programs see Janet C. Gornick (1999). “Paid family leave” refers to policy provisions that grant rights and benefits for various breaks from paid employment. These include, for example, maternity leave, paternity leave, parental leave, sick-child leave, and short intermittent leaves—sometimes referred to as leave for family reasons—that allow parents to respond to nonroutine caregiving demands. The United States is one of five countries in the entire world that do not have a national law granting paid maternity leave. It is one of just a few industrialized countries that do not have a national policy that provides paid leaves for both mothers and fathers. For a detailed discussion of this point, see Rebecca Ray, Janet C. Gornick, and John Schmitt (2008). Although space limitations prevent us from discussing them here, each of these countries also provides various kinds of “leave for family reasons.” These leaves grant mothers and fathers time off throughout their children’s lives to attend to short-term and unexpected needs. See Gornick and Meyers (2003) for details. For example, mothers in Finland may collect a low flat-rate benefit (a “home care” benefit) for about two years following the end of maternity and parental leave, until the child’s third birthday. The benefit is allowed only if the child is not in public child care. Parents may also choose to use that payment to purchase care from a private child-care provider. Although Norway also provides extensive public ECEC, the costs of this care fall much more heavily on parents (because of high co-payments), and supply shortages have contributed to the growth of a “black market” in private, unregulated care arrangements. Emsellem (2006) also notes that when unemployment benefits are available, many employers benefit as well because they are able to preserve their workforce during periods of temporary layoffs. The details underlying these cost estimates are available from the authors. Cost estimates for paid family leave are based on estimates for the California program by Arindrajit Dube and Ethan Kaplan (2002), adjusted upward to reflect a longer benefit period and higher wage replacement rate. Arindrajit Dube (personal communication, May 6, 2004) provided helpful suggestions for this estimate. Estimates for earlychildhood education and care costs are based on Susan W. Helburn and Barbara

/

287

Making the Work-Based Safety Net Work Better Bergmann (2000), adjusted for the number of children receiving care on a part- or fulltime basis, and increased by about $2,000 per child annually to improve providers’ salaries and service quality. The resulting per-child estimates are similar to the actual price of high-quality preschool services currently provided in public settings. 15. The estimated 2.5 percent of GDP in Sweden and 1.5 percent of GDP in the U.S. represent roughly equivalent levels of expenditures per capita because per capita GDP is higher in the United States than in Sweden. See Gornick and Meyers (2003) for a more extended discussion.

REFERENCES Acs, Gregory, Harry J. Holzer, and Austin Nichols. 2005. How Have Households with Children Fared in the Job Market Downturn? New Federalism Issues and Options for the States, Series A, no. A-67. Washington, D.C.: Urban Institute. Adema, Willem, and Maxime Ladaique. 2005. Net Social Expenditure, 2005 Edition: More Comprehensive Measures of Social Support. OECD Social, Employment and Migration Working Papers, no. 29. Paris: Organisation for Economic Co-Operation and Development. Barnett, W. Steven, Jason T. Hustedt, Kenneth B. Robin, and Karen L. Schulman. 2005. The State of Preschool: 2005 State Preschool Yearbook. New Brunswick, N.J.: National Institute for Early Education Research. Blank, Rebecca M., Sheldon H. Danziger, and Robert F. Schoeni. 2006. “Work and Poverty During the Past Quarter Century.” In Working and Poor: How Economic and Policy Changes Are Affecting Low Wage Workers, edited by Rebecca M. Blank, Sheldon H. Danziger, and Robert F. Schoeni. New York: Russell Sage Foundation. Blau, David, and Janet Currie. 2004. “Preschool, Day Care, and Afterschool Care: Who’s Minding the Kids?” NBER Working Paper no. W10670. Cambridge, Mass.: National Bureau of Economic Research. Burman, Leonard E., and Laura Wheaton. 2005. “Who Gets the Child Tax Credit?” Report. Washington, D.C.: Urban Institute. Available at: http://www.urban.org/Uploaded PDF/411232_child_tax_credit.pdf (accessed on March 23, 2008). Cauthen, Nancy. 2005. Whose Security? What Social Security Means to Children and Families. New York: Columbia University, National Center for Children in Poverty. Currie, Janet. 2006. The Invisible Safety Net: Protecting the Nation’s Poor Children and Families. Princeton, N.J.: Princeton University Press. Dodson, Lisa, and Ellen Bravo. 2003. “Managing Work and Family: Why It’s Not Working for the Bottom Third.” In Work, Family and Democracy, edited by Christopher Beem and Jody Heyman. Princeton, N.J.: Princeton University Press. Dube, Arindrajit, and Ethan Kaplan. 2002. “Paid Family Leave in California: An Analysis of Costs and Benefits.” Unpublished manuscript. Berkeley, Calif.: Labor Project for Working Families. Emsellem, Maurice. 2006. “Innovative State Reforms Shape New National Economic Security Plan for the 21st Century.” Special Report. New York: National Employment Law

288

/

Beyond the Safety Net Project. Available at: http://www.nelp.org/site/publications/P200 (accessed on April 20, 2009) Figart, Deborah M., and Lonnie Golden. 1998. “The Social Economics of Work Time: An Introduction.” Review of Social Economy 56(4): 411–24. Garfinkel, Irwin, Lee Rainwater, and Timothy Smeeding. 2006. “A Re-Examination of Welfare State and Inequality in Rich Nations: How In-Kind Transfers and Indirect Taxes Change the Story.” Journal of Policy Analysis and Management 25(4): 897–918. Gatenio, Shirley G., and Sheila B. Kamerman. 2006. “Investing in Children: Public Commitment in Twenty-One Industrialized Countries.” Social Service Review 80(2): 239–83. Giannarelli, Linda, Sarah Adelman, and Stefanie Schmidt. 2003. “Getting Help with Child Care Expenses.” Assessing the New Federalism Project, Occasional Paper no. 62. Washington, D.C.: Urban Institute. Giannarelli, Linda, and James Barsimantov. 2000. “Child Care Expenses of American Families.” Assessing the New Federalism Project, Occasional Paper no. 40. Washington, D.C.: Urban Institute. Gornick, Janet C., and Marcia K. Meyers. 2003. Families That Work: Policies for Reconciling Parenthood and Employment. New York: Russell Sage Foundation. ———. 1999. “Income Maintenance and Employment Supports for Former Welfare Recipients: The United States in Cross National Perspective.” In Rethinking Income Support for the Working Poor, edited by Evelyn Ganzglass, and Karen Glass. Washington, D.C.: National Governors’ Association. Gould, Elise. 2006. “Health Insurance Eroding for Working Families.” EPI Briefing Paper no. 175. Washington, D.C.: Economic Policy Institute. Available at: http://www.epi .org/content.cfm/bp175 (accessed on March 23, 2008). Hacker, Jacob S. 2002. The Divided Welfare State: The Battle over Public and Private Social Benefits in the United States. Cambridge; New York: Cambridge University Press. Helburn, Susan W., and Barbara Bergmann. 2002. America’s Child Care Problem. New York: St. Martin’s Press. Heyman, Jody 2000. The Widening Gap: Why America’s Working Families Are in Jeopardy—and What Can Be Done About It. New York: Basic Books. Holt, Stephen D. 2006. The Earned Income Tax Credit at Age 30: What We Know. Washington, D.C.: Brookings Institution. Holt, Stephen D., and Jennifer L. Romich. 2007. “Marginal Tax Rates Facing Low- and Moderate-Income Workers Who Participate in Means-Tested Programs.” National Tax Journal 60(2): 253–276. Howard, Christopher. 1997. The Hidden Welfare State. Princeton, N.J.: Princeton University Press. Magnuson, Katherine A., Marcia K. Meyers, and Jane Waldfogel. 2007. “Public Funding and Enrollment in Formal Child Care in the 1990s.” Social Service Review 81(1): 47–83. Meyers, Marcia K. 2007. “The Institutional Architecture of Anti-Poverty Policy in the United States: Looking Back, Looking Ahead.” Focus, Spring–Summer, 58–61. Meyers, Marcia K., and Sandra Garcia. 2006. “The Social benefits package for New York City Families.” Unpublished manuscript. Seattle: University of Washington.

/

289

Making the Work-Based Safety Net Work Better Meyers, Marcia K., Janet C. Gornick, Laura R. Peck, and Sarah Bruch. 2008. “Changes in the United States Social benefits package at the State Level, 1996 to 2005.” Unpublished manuscript. Seattle: University of Washington. Minoff, Elisa. 2006. The UK Commitment: Ending Child Poverty by 2020. Washington, D.C.: Center for Law and Social Policy. Mishel, Lawrence, Jared Bernstein, and Sylvia Allegretto. 2007. The State of Working America 2006–2007. Washington, D.C.: Economic Policy Institute. Needles, Karen, and Walter Nicholson. 1999. An Analysis of Unemployment Insurance Durations Since the 1990–1992 Recession. Princeton, N.J.: Mathematica Policy Research. Organisation for Economic Co-Operation and Development (OECD). 2001. “Balancing Work and Family Life: Helping Parents Into Paid Employment.” Employment Outlook 2001(June): 129–66. ———. 2007a. Babies and Bosses: Reconciling Work and Family Life; A Synthesis of Findings for OECD Countries. Paris. ———. 2007b. Employment Outlook, Statistical Annex. Paris. Peattie, Lisa, and Martin Rein. 1983. Women’s Claims: A Study in Political Economy. London: Oxford University Press. Phillips, Katherin Ross. 2004. Getting Time Off: Access to Leave Among Working Parents. New Federalism, National Survey of America’s Families, Series B, no. B-57. Washington, D.C.: Urban Institute. Rainwater, Lee, Martin Rein, and Joseph Schwartz. 1986. Income Packaging in the Welfare State: A Comparative Study of Family Income. London: Oxford University Press. Rainwater, Lee, and Timothy Smeeding. 2003. Poor Kids in a Rich Country: America’s Children in Comparative Perspective. New York: Russell Sage Foundation. Ray, Rebecca, Janet C. Gornick, and John Schmitt. 2008. Parental Leave Policy in 21 Countries: United States Parents Left Out. Washington, D.C.: Center for Economic Policy Research. Remler, Dahlia K., and Sherry A. Glied. 2003. “What Other Programs Can Teach Us: Increasing Participation in Health Insurance Programs.” American Journal of Public Health 93(1): 67–74. Romich, Jennifer L., Jennifer Simmelick, and Stephen D. Holt. 2007. “When Working Harder Does Not Pay: Low Income Working Families, Tax Liabilities, and Benefit Reductions.” Unpublished manuscript. Seattle: University of Washington. Smith, Kristen E., Barbara Downs, and Martin O’Connell. 2001. Maternity Leave and Employment Patterns: 1961–1995. Washington, D.C.: United States Bureau of the Census. Smeeding, Timothy. 2004. “Public Policy and Economic Inequality: The United States in Comparative Perspective.” Paper presented at the Inequality and American Politics Conference. Campbell Institute Seminar, Syracuse University, Syracuse (February 20). Stone, Chad, Robert Greenstein, and Martha Coven. 2007. Addressing Longstanding Gaps in Unemployment Insurance Coverage. Washington, D.C.: Center on Budget and Policy Priorities. United States Department of Health and Human Services. 2000. Access to Childcare for LowIncome Working Families. Washington, D.C.: Administration for Children and Families. United States Department of Labor. 2000. Balancing the Needs of Families and Employers: The Family and Medical Leave Surveys. 2000 Update. Available at: http://www.dol.gov/asp/ fmla/main2000.htm (accessed on July 9, 2008).

290

/

Beyond the Safety Net ———. 2008. “Coverage Status of Workers Under Employment Provided Pension Plans.” Washington, D.C.: Employee Benefits Security Administration. Available at: http:// www.dol.gov/ebsa/publications/99pensionreport.html (accessed on March 23, 2008). Vroman, Wayne. 2003. Extending Unemployment Insurance (UI) Protection. Washington, D.C.: Urban Institute. Waldfogel, Jane. 1999. “Family Leave Coverage in the 1990s.” Monthly Labor Review 122(10): 13–23. ———. 2007. Welfare Reforms and Child Wellbeing in the US and UK. Working paper 2008:11. Uppsala, Sweden: Institute for Labour Market Policy Evaluation. Whiteford, Peter, and Willem Adema. 2007. What Works Best in Reducing Child Poverty: A Benefit or Work Strategy? OECD Social, Employment and Migration Working Papers no. 51. Paris. Williams, Joan. 2006. One Sick Child Away from Being Fired: When “Opting Out” Is Not an Option. San Francisco: University of California, Hastings College of the Law, Center for WorkLife Law. Zedlewski, Sheila, and Seth Zimmerman. 2007. Trends in Supports for Low Income Families with Children. Washington, D.C.: Urban Institute.

/

291

Chapter 9 The Politics of Low-Income Families in the United States R. Kent Weaver

P

olicies for low-income families are an area of great controversy in the United States. Although “greater self-sufficiency” is almost universally seen as a desirable goal for low-income families, there is much less agreement on what “self-sufficiency” means, what the barriers are to achieving it, and what role government should play in helping families achieve it. Disagreements about how to improve the life prospects of low-income families often take place along partisan and ideological lines. In the short term, the election of Barack Obama to the U.S. presidency in 2008, with increased Democratic majorities in both chambers of Congress, has created an opening for major new initiatives in policies that affect low-income families. Over the long run, however, the financial market crisis that coincided with and facilitated the electoral success of Obama and his fellow Democrats, as well as the prospect of a prolonged economic downturn, can be expected to have a much more mixed impact: they may increase the public appetite for bold policy action, but they also tighten the fiscal vice constraining policy initiatives that carry big price tags. This chapter examines the potential for substantial change in policies that affect the economic position and life chances of low-income families by examining the political and fiscal constraints that have shaped these policies in the past and are likely to continue to shape policy initiatives over the next two decades. I focus here on policies that are consistent with the range of approaches taken by individual authors in this volume—that is, that seek to move families out of poverty by promoting human capital development, work, and labor-market earnings but that also recognize that a substantial public role in providing income transfers and services will also be required to lower family poverty.1 The starting point of my analysis is a recognition that policies intended to help low-income families will do

292

/

The Politics of Low-Income Families

little good if they cannot be enacted and implemented. And even if policy analysts could tell politicians definitively which social policy interventions would improve the lives of these families most effectively and most efficiently, there is no guarantee that politicians would follow their advice. Low-income people in the United States have a number of weaknesses when it comes to influencing public policy choices. In terms of political resources, they lack money to make campaign contributions; they are poorly organized to represent their collective interests; their voting levels and other rates of political participation are generally low; and they are rarely “median voters” whose votes both major political parties hope to capture. Thus, as Hugh Heclo (1994) puts it, “Antipoverty policies are less a matter of demands poor people make in the political process and more a function of what other people decide to do to and for them” (397; see also Soss 1999; Mettler and Soss 2004). Moreover, the mixed nature of low-income families—children who are viewed sympathetically by the public, and adults who (as we will see) are viewed with much more ambivalence—means that policies to aid these families may also be viewed by the public with ambivalence, or worse. The lack of political power of low-income families and uncertain public support for many policies benefiting those families have important impacts on the choices made by federal and subnational governments. I examine the constraints on policy toward low-income families in three stages. First, I very briefly outline the major contours and evolution of current U.S. policies to assist low-income families. I discuss several different broad types of programs—I call them policy “streams”—such as social insurance, refundable tax credits, and both cash and inkind means-tested assistance for low-income families, as well as the distinctive political dynamics of each policy stream. I argue that there is substantial variation in political dynamics across programs and over time, not only across but also within each of these political streams. Nevertheless, much can be learned by examining the distinctive political dynamics of these political streams. Much of the variation across and within each stream of programs can be explained by a series of enduring political constraints that characterize American politics. These constraints include public opinion toward low-income families and race, gridlock-prone political institutions, increasing fiscal pressures, and federalism. These constraints are the focus of the middle section of the chapter. The third section of the chapter examines the implications of these constraints for the policy and strategic options pursued by advocates for low-income families. In focusing on multiple policy streams my interest is in thinking through the conditions that facilitate and inhibit innovation in and expansion and retrenchment of programs serving low-income families; only after we have done that can we begin to think through options for future political action. If, for example, social insurance offers major potential advantages for promoting a work-based strategy for low-income American families, but the conditions expanding social insurance programs to those families are hardly ever present, then advocates for the poor and their political allies need to focus on alternative policy streams, or they are likely to have a long wait to achieve progress toward their goals.

/

293

Making the Work-Based Safety Net Work Better

POLICY STREAMS U.S. policies toward low-income families are incredibly complex. A recent Congressional Research Service study lists eighty-five income-tested programs, many of which provide some benefits to low-income families with children (Burke 2003). Another sixty tax-expenditure programs help to finance education, training, social services, health, and income security (U.S. Office of Management and Budget 2007, 288–89). In addition, a variety of other public policies, such as minimum-wage laws, affirmative action policies, and policies toward education, may affect the economic prospects of low-income families. Rather than a conventional chronological history of specific programs (for excellent examples, see Patterson 1986; Katz 1987; Heclo 1986), or policy toward antipoverty programs generally, I will focus here on several political “streams” of social policy for low-income families—different types of policies with distinctive political dynamics of program initiation, maintenance and growth over time, and resistance to retrenchment or termination initiatives. Policymaking for lowincome families in some streams is marked by pronounced ebbs and flows—a few torrents of policymaking and some serious droughts. Moreover, support for some programs may dry up almost entirely over time. This happened in the United States, first with direct public employment programs and later with open-ended cash assistance for single-parent families that is not conditioned on work.2 Four caveats are important at the outset. First, the search for commonalities over time and across programs inevitably lumps together historical events and movements that have much that is not common, and misses much programspecific nuance that may be important in explaining the evolution of individual programs. Second, I will focus primarily on changes in policy that are enacted by elected politicians rather than by the courts (Melnick 1994) or by bureaucrats in the implementation process. The increasingly conservative direction of recent Supreme Court decisions suggests that the courts are not likely to be as promising a venue for an expanded commitment to the poor in the near future as they were in the 1960s and early 1970s, unless there is a substantial turnover in the Court’s membership. And there is no question that the way that clients are treated in employment and welfare offices (Fording, Soss, and Schram 2007; Riccucci 2005; Keiser, Mueser, and Choi 2004) and by Medicaid providers affects the way that policy toward low-income people is felt on the ground. But both litigation-based strategies and implementation issues remain largely beyond the scope of this chapter. Third, although I try to incorporate a broad historical scope in the essay, I focus primarily on events since 1990. Focusing primarily on recent events makes most sense if there are secular trends in American politics; of course, if there are “long cycles” or swings of the pendulum, and the United States has been near one end for the past two decades, focusing disproportionately on the recent past may underestimate the range of political possibilities. Finally, it is important to look not just at the initial creation of programs but also at their evolution over time to avoid overconcentration on the few “big bangs” of program creation (Leman

294

/

The Politics of Low-Income Families

1977) and termination rather than on more mundane politics that shapes programs’ expansion and contraction. Figure 9.1 provides a time series of federal and many state expenditures on selected categories of programs that provide substantial cash and in-kind benefits to low-income families, with the expenditures shown as percentages of GDP. (See notes to the figure for the details of how expenditures are calculated. Where possible, I have tried to separate out the share of expenditures that are directed to lowincome families from other groups such as elderly households without children— and only include the former). Table 9.1 provides a snapshot of the data in two years, 1980 and 2005, in percentage terms. There are several striking patterns in the data. First, although there has been a major increase in expenditures for low-income families over this period in real dollar terms, expenditures on these programs as a percentage of GDP have not increased substantially except in the area of health care (figure 9.1). Second, there has been a very substantial shift in the expenditures across categories (table 9.1). There has been a huge increase in the role of refundable payments under the Earned Income Tax Credit and Child Tax Credit (the latter did not exist in 1980), and in health expenditures (under S-CHIP [State Children’s Health Insurance program] and Medicaid). The relative share taken by cash assistance under Aid to Families with Dependent Children (AFDC) and its successor program, Temporary Assistance to Needy Families (TANF), on the other hand, has declined precipitously, and the increase in tax-based benefits roughly offsets the sharp drop in AFDC and TANF assistance. Put another way, aid to low-income families predicated on work has supplanted assistance to families where participation in paid work is not the norm. The role of Unemployment Insurance has also declined. Expenditure patterns, in any case, tell only part of the story. To understand how the United States arrived at these policy choices, and where they are likely to go in the future, it is necessary to look in more detail at individual policy streams.

Social Insurance: Stepping Stone or Firewall? Social insurance programs—compulsory, contributory, payroll-tax-financed programs that serve broad swaths of the population on the basis of their contributions rather than on need—are the core of the American welfare state. (Except for Unemployment Insurance, these programs are not shown in figure 9.1 and table 9.1, since it is hard to separate out which expenditures go to low-income families.) Many policy gains for the poor have been won not by meeting their concerns directly but by, as Heclo put it, “incorporating the poor through the political backdoor” (1986, 315) in social insurance programs, through which the poor gained disproportionately. The United States stands out from most other wealthy countries in not having a universal (“demogrant”) or quasi-universal system of child allowances that provide an income floor for families with children (Kamerman and Gatenio 2002). And low-income families generally benefit less from social in-

/

295

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

2.25

2.50

2.75

72

19

FIGURE 9.1

Percentage of GDP

74 19

78

19

80

19

82

19

84

19

86

19

88

19

Year

90 19

92 19

AFDC/TANF Cash Assistance Tax-Based Foster Care, Child Care, TANF Noncash, and LIHEAP Food Stamps

76 19

94 19

96

19

98

20

00

20

02

20 Nutrition Except Food Stamps Health Care Unemployment Insurance

19

/ Spending on Low-Income Families in Selected Programs, as a Percentage of GDP

04

20

06

“Tax-based” payments are the refundable section only of the federal Earned Income Tax Credit and the Child Tax Credit. Data are from Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3. “Foster care, child care, TANF noncash, and LIHEAP” data are federal expenditures only for child care and foster care plus LIHEAP plus combined federal and state expenditures under TANF other than expenditures under basic assistance and expenditures under prior law. Sources are Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3, and U.S. Department of Health and Human Services, Administration for Children and Families, “TANF Financial Data,” Table F, various years. “Food Stamps” are federal expenditures for all categories of Food Stamp recipients, and “Nutrition Except Food Stamps” are federal expenditures under the child nutrition and special milk programs and the supplemental feeding programs (WIC and CSFP programs). Source is Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3. Healthcare expenditures are federal and state vendor payments under the Medicaid program for dependent children under age twenty-one and for adults in families with dependent children, plus federal payments under the State Children’s Health Insurance Program. Sources are for Medicaid, Social Security Bulletin, Annual Statistical Supplement, 1984–85, Table 155, and Social Security Bulletin, Annual Statistical Supplement, 2006, Table 8E; for S-CHIP, Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3. Medicaid expenditures after 2003 are extrapolated from 2003 assuming a constant ratio between federal Medicaid expenditures and total Medicaid expenditures for children under age twenty-one and adults in families with dependent children. Unemployment Insurance expenditures are total UI expenditures as listed in Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3.

Sources: “AFDC/TANF Cash Assistance” includes both federal and state expenditures. Data are from House and Ways Committee, 1998 Green Book, p. 411, “Total Benefit Payments”), and the U.S. Department of Health and Human Services, Administration for Children and Families, “TANF Financial Data”, Table F, various years. Expenditures for 2006 and 2007 are extrapolated from 2005 data, assuming a constant relationship to the 2005 federal block grant.

Making the Work-Based Safety Net Work Better TABLE 9.1

/ Spending on Low-Income Families in Selected Programs, 1980 and 2005 (as a Percentage of Total)

Program Category AFDC/TANF cash assistance Foster care, child care, TANF noncash, and Low Income Home Energy Assistance Program Tax-based assistance Food Stamps Nutrition except Food Stamps Health care Unemployment Insurance

1980

2005

23

5

3 2 18 9 12 33

6 21 14 8 32 14

Source: “AFDC/TANF Cash assistance” includes both federal and state expenditures. Data are from House and Ways Committee, 1998 Green Book, p. 411, “Total Benefit Payments”), and the U.S. Department of Health and Human Services, Administration for Children and Families, “TANF Financial Data”, Table F, various years. Expenditures for 2006 and 2007 are extrapolated from 2005 data, assuming a constant relationship to the 2005 federal block grant. “Tax-based” payments are the refundable section only of the federal Earned Income Tax Credit and the Child Tax Credit. Data are from Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3. “Foster care, Child Care, TANF noncash, and LIHEAP” data are federal expenditures only for child care and foster care plus LIHEAP plus combined federal and state expenditures under TANF other than expenditures under basic assistance and expenditures under prior law. Sources are Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3, and U.S. Department of Health and Human Services, Administration for Children and Families, “TANF Financial Data,” Table F, various years. “Food Stamps” are federal expenditures for all categories of Food Stamp recipients, and “Nutrition except Food Stamps” are federal expenditures under the child nutrition and special milk programs and the supplemental feeding programs (WIC and CSFP programs). Source is Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3. Health-care expenditures are federal and state vendor payments under the Medicaid program for dependent children under age twenty-one and for adults in families with dependent children, plus federal payments under the State Children’s Health Insurance Program. Sources are for Medicaid, Social Security Bulletin, Annual Statistical Supplement, 1984–85, Table 155, and Social Security Bulletin, Annual Statistical Supplement, 2006, Table 8E; for S-CHIP, Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3. Medicaid expenditures after 2003 are extrapolated from 2003 assuming a constant ratio between federal Medicaid expenditures and total Medicaid expenditures for children under age twenty-one and adults in families with dependent children. Unemployment Insurance expenditures are total UI expenditures as listed in Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, Table 11-3.

surance than from universal demogrants, since social insurance usually provides little to those with low career earnings and brief or interrupted work histories. The conventional wisdom on social insurance programs is that while they may face major obstacles to winning enactment, once enacted they are politically resilient and prone to expansion: the fact that benefits are based on contributions gives beneficiaries both a moral and legal claim on benefits owed. In addition, the

298

/

The Politics of Low-Income Families

broad distribution of social insurance benefits across economic classes makes them prone to electoral bidding wars during relatively flush economic times and heavily resistant to cutbacks when times are tough. If there is a reasonable amount of fiscal slack, and a program remains uncontroversial (no obvious fraud or abuse), there may be opportunities for program enhancements, usually of relatively modest size, that benefit program clientele in general and in particular “tuck the poor away” and provide them with disproportionate benefits. The classic example of this is the incremental expansion of the Social Security program in the United States, which in its early years benefited from a dominant role played by a supportive “policy cartel” of program executives, policy experts, and congressional oversight committees who made a series of incremental, expansionary policy changes that often gained very little public attention. Wilbur Cohen, one of the primary architects of Social Security’s growth, described his philosophy of expanding social insurance as follows (Derthick 1979, 26): Take a piece of salami and slice it very thin and pile slice upon slice, so that eventually you have a very good sandwich. And that is my concept of the evolution of social legislation; take a bite at the time and digest it, and then go on to the next phase in an orderly evolution that is acceptable by [sic] the body politic as being practical realistic, and one they’re willing to build upon.

Recognizing the political shakiness of a strategy based on programs targeted to the poor, scholars such as Theda Skocpol (1991) and William Julius Wilson (1987) have argued for a pursuit of a political strategy that Skocpol (1991) calls “targeting within universalism”—or more accurately, “targeting within social insurance”—because, as noted, broad-based income transfer and health insurance programs in the United States have been based on social insurance rather than universal demogrant principles.3 While there is much truth to the conventional wisdom on the political advantages of social insurance, several caveats are also in order in applying this wisdom to low-income families. The most obvious is the connection of social insurance to contributions, and thus to steady workforce participation—a connection that is tenuous among many low-income families in the United States. Moreover, if the major political advantage of a strategy of targeting within social insurance is that benefits are distributed to a broad constituency and thus are likely to be broadly popular, the flip side (and consequent disadvantage) is that these programs are likely to be extremely expensive, and new programs are likely to intrude upon existing arrangements and their providers (for example, health insurance companies). Thus they are likely to spark opposition from political conservatives and from the business community, which worries that increases in both corporate income taxes and payroll taxes will be needed to finance social insurance programs with very broad clienteles. These complex dynamics are visible in the three largest social insurance programs in the United States: Social Security, Medicare, and Unemployment Insurance. Social Security (technically, Old Age, Survivors, and Disability Insurance, or

/

299

Making the Work-Based Safety Net Work Better

OASDI) provides the closest fit to the conventional wisdom: a program that initially was narrowly drawn was expanded over time. The political processes that led to “backdoor” benefits to low-income families were often complex and drawn out. Social Security originally excluded many of the poorest workers (domestic workers and agricultural laborers), and did not provide benefits to survivors and the disabled, but these groups were increasingly included (survivors in 1939, the disabled in 1956) as Social Security was expanded over time. But as Heclo (1986) has noted about the New Deal, “Few political resources were expended in trying to legitimize any larger norms of redistribution”—in fact, the public emphasis of program managers was on the principle that benefits were earned rather than on the low-profile redistribution (Derthick 1979). The Social Security benefit formula benefits former low-income workers by providing both retirees and the disabled with a higher replacement rate on the first dollars of earned income. Although Social Security is thought of primarily as a program for seniors and the disabled, it also provides benefits to surviving children of deceased workers and to families of disabled workers.4 Indeed, the Center on Budget and Policy Priorities estimated in 2005 that Social Security does more to reduce the child poverty gap—”the cumulative dollar amount by which incomes of all poor families with children fall short of the poverty line“—than any other program.5 Of course, Social Security is a vastly larger program than the others. It should also be recognized that because of its link to past parental earnings, Social Security provides less of a safety net for low-income families than a flat-rate program would, even though Social Security does have a progressive benefit formula (the replacement rate for low-earning individuals or couples is higher than the replacement rate for high-earning individuals or couples).6 Moreover, the expansionist story does not fit the Social Security experience of the past thirty years, which is instead a story of very serious funding crises in 1977 and from 1981 to 1983, resulting in retrenchment, followed by a quarter century of a virtual freeze in policy as a result of partisan stalemate between Democrats, who support the status quo, and Republicans, who favor major restructuring and adamantly oppose any additional payroll taxes (Weaver 2005a, 2005b; Edwards 2008, chapters 7 and 8). The prolonged stalemate has meant that potential reforms to the program that could aid low-income families—for example, providing Social Security credits for women who reduce their participation in the paid labor market while raising young children, as has been enacted in many European countries—have simply not been seriously considered in the United States. Nor are they likely to be considered in the near future, given the financial prospects for Social Security, except possibly as a potential “sweetener” in a major Social Security reform and retrenchment package designed to handle long-term funding shortfalls in the program. Other social insurance programs in the United States have also had major limitations as vehicles for aiding low-income families. The boundaries of the Medicare program clientele have remained relatively fixed rather than serving as a launching pad for expansion into a national health insurance program that would serve all low-income families.7 Expansion of public health insurance has

300

/

The Politics of Low-Income Families

intermittently been on the agenda in the United States, and much ink and paper has been utilized explaining the failure of national health insurance initiatives by the Truman, Carter, and Clinton administrations. The role of the assassin has been variously played by business interests; a public that was only superficially committed to universal provision, more concerned about a decline in quality of its own care, and susceptible to manipulation by insurance company lobbying; a political culture that was suspicious of a large government role in social provision; incompetent handling by the sponsoring presidential administration; fragmented political institutions; poor timing due to fiscal or inflationary pressures; and a number of other “usual suspects” (Steinmo and Watts 1995; Peterson 1998; Quadagno 2005; Gordon 2004). Most of these explanations are plausible, and the failure of any of the major reform initiatives to come close to congressional passage suggests that the failure of health-care reform had multiple causes and was overdetermined. But another component of the causal mix of constraints on efforts to enact a comprehensive national health insurance package since 1965 is the existence of Medicare itself. Public demand for a comprehensive approach to health-care reform has been weakened because Medicare provides health insurance to the elderly and disabled, clienteles that (along with the poor, served by Medicaid) are least likely to be able to obtain affordable health insurance in private markets. In short, the history of Medicare and Medicaid over the past fifty years has largely followed the original conception of Wilbur Mills (chairman of the House Ways and Means Committee when the two programs were created) that Medicare and Medicaid would constitute a fence (or firewall) against a more comprehensive national health insurance program (Marmor 1973, 79). Democratic presidential candidates in 2008, including the eventual winner, Barack Obama, for the most part took more incremental approaches than those proposed by Democratic presidents in the last half of the twentieth century. They focused on individual mandates, subsidies to help low-income people purchase health insurance, and marginal expansion of existing public programs rather than bolder types of reforms (Lewin Group 2008).8 The third major social insurance program with the potential to affect lowincome families is Unemployment Insurance. The politics of Unemployment Insurance have been heavily influenced by two major programmatic features. First, it is financed entirely by a tax on employers, and the effective rate of state taxation is not fixed. This increases the attentiveness of employers to the program and focuses their efforts on restraining effective payroll tax rates. A second major feature of Unemployment Insurance is its federal character, to be discussed further in the second section of this chapter. Moreover, the grounds for receipt of Unemployment Insurance have remained narrowly drawn rather than expanding over time in ways that would benefit low-income families. For example, inclusion of maternity as a grounds for receipt of employment insurance, as is done in Canada (see Pal 1985), has never really made it on to the agenda in the United States. A very limited right to unpaid parental leave was established in federal law in 1993, without a connection to Unemployment Insurance (Elving 1995).

/

301

Making the Work-Based Safety Net Work Better

Overall, the U.S. experience with social insurance programs offers several lessons about the prospects for using those programs as a vehicle for aiding lowincome families. Social insurance programs do tend to generate strong supportive coalitions, as political scientists have long noted (Pierson 1994), but these coalitions will not necessarily support an extension of benefits to new constituencies. However, recipient groups, notably the gigantic AARP, are very well placed to resist policy initiatives that attempt to impose losses on all or part of its constituency. Thus the constellation of interest pressures militates strongly against major changes in most mature social insurance programs (Oberlander 2003, chapter 3). When benefits have been enhanced in mature social insurance programs, as with the addition of prescription drug benefits to Medicare in 2003, it has generally gone to existing, well-organized clienteles who can at least partially pay for their own benefits. Efforts to enact massive cross-subsidies to finance benefit expansions in contributory programs, as was clearly the case with Medicare catastrophic insurance, and arguably the case with the Clinton health-care proposal, have generally been political disasters. Social insurance programs do tend to be resistant to general pressures for fiscal constraint. But mature social insurance programs also appear to offer limited opportunities for aiding new constituencies such as low-income families once they are mature, especially when, as is the case in the United States, a politically well-entrenched force adamantly opposes any increases in payroll tax—as has been the case with congressional Republicans since 1983. Moreover, they are susceptible to “trust fund crises”—projections that their revenue sources will not be adequate to cover anticipated expenditures in the immediate (Social Security in 1977 and 1983) or intermediate (Medicare Part A) terms (see Oberlander 2003, chapter 4). Payroll tax financing and the threat of recurrent trust fund crises not only constrain expansion, they also prompt retrenchment. This retrenchment is often disguised (for example, cuts in provider reimbursement rates within Medicare) or stretched out in time (increases in the Social Security retirement age), but it is retrenchment nonetheless.

Income-Tested Cash Assistance Income-tested cash assistance programs have two major advantages for delivering benefits to low-income families: they are less expensive at providing an income floor than universal programs, because expenditures are targeted at those with low incomes, and unlike in-kind programs, they allow recipients with limited resources to choose the expenditures that matter the most to them rather than having those choices made for them.9 Weighed against these advantages, however, are a number of disadvantages. Income tests, which in the United States are usually accompanied by assets tests, are frequently complex and costly to administer. Both the stigma and the administrative hassles associated with initial certification and recertification may interfere with program take-up by people who are eligible for the program. And benefit phase-outs as income increases may represent significant work disincentives.

302

/

The Politics of Low-Income Families

The primary problems of income-tested cash assistance for low-income families in the United States have primarily been political, however. The Social Security Act of 1935 created federal-state shared-cost programs for four categories of individuals or households who were not expected to work: the aged, blind, disabled, and families where the (presumably male) breadwinner was absent. The first three categories were largely nationalized as the Supplemental Security Income program in 1972 (Derthick 1990; Pierson 1995), although benefits remained very low, and assets tests are severe. But the Aid to Families with Dependent Children (originally Aid to Dependent Children) program remained a shared federal-state program, and a program mired in controversy. The federal-state structure of AFDC gave substantial leeway over benefit levels, eligibility standards, and program administration to the states, and southern states tended to be more conservative on all of these criteria than most states in other regions.10 Originally intended as a program to assist widows, AFDC increasingly came to serve families where the mother was divorced from, deserted by, or had never been married to the father. It also grew to overrepresent African Americans (relative to the overall population, if not the low-income population). In addition, public expectations about whether the women heading these households should be expected to work (especially if their children were of school age), were always ambiguous at best, and work expectations grew stronger as more middle-class women entered the workforce. AFDC had other political problems as well. Its inclusion as part of the Social Security Act of 1935 also meant that AFDC was under the jurisdiction of the House Ways and Means Committee. It was a low-priority item for a powerful (and through the 1960s consensually-oriented) body that had more “important” issues like taxes, trade, and Social Security to deal with, and did not view AFDC as a potential starting block for an integrated family policy.11 Given these constraints, it is not surprising that AFDC experienced little of the incremental program expansion through a combination of “insider politics” and electoral competition that enhanced Social Security in its first thirty years. Legislative expansions of AFDC were infrequent and modest,12 and its politics involved contentious and divisive “tireless tinkering” for an unpopular program and clientele.13 Indeed, the huge growth of AFDC rolls and expenditures in the 1960s is largely attributed to court cases that struck down a number of state-imposed restrictions on eligibility, as well as higher program take-up resulting from decreased program stigma and a short-lived welfare rights movement (Heclo 2001). The Nixon administration’s initiative to convert AFDC into a more nationally uniform negative income tax program called the Family Assistance Plan foundered on both cost projections and suspicions of its possible work disincentives and incentives for childbearing (Moynihan 1973). A similar effort by President Carter in 1977 met the same fate (Lynn and Whitman 1981). AFDC policy was increasingly tasked with the very difficult assignment of engaging people who had become disengaged from the workforce, or who had never been engaged in the workforce, rather than the much simpler task of writing checks. By 1967 Congress was enacting legislation intend to curtail AFDC ex-

/

303

Making the Work-Based Safety Net Work Better

penditures and take-up, and to encourage AFDC recipients to work (Melnick 1994). Efforts to increase work among AFDC recipients took various forms over the next thirty years, including increasing (1967), and decreasing (1981) earnings disregards for AFDC recipients who entered work, adding (initially toothless) work requirements, and providing funding for training, child care, and other services to facilitate transitions into work (1988). Critics of the program, most notably Charles Murray in his 1984 book Losing Ground, portrayed AFDC as the problem rather than the solution to what was now labeled “welfare dependency”—not just for discouraging work but also for encouraging nonmarital births. In the early 1990s, states were given increased leeway to try new approaches to encourage or require work among welfare mothers, ranging from stiffer work requirements to “family caps” (no additional benefit when a welfare recipient had more children) and increased earnings disregards and extensions of Medicaid coverage to those leaving the AFDC rolls. Incremental reforms for AFDC came to an end in 1996, when a new Republican majority in Congress, emboldened by Bill Clinton’s ambiguous campaign pledge to “end welfare as we know it,” enacted legislation quite different from the Clinton administration’s initial 1994 proposals: the AFDC program was converted into a block grant to the states labeled Temporary Assistance to Needy Families (TANF). Under the new TANF program, individual entitlement to benefits was ended, and the amount received by states was essentially fixed rather than fluctuating with caseloads. States received far more discretion in some important ways, notably freedom to spend their block grant funds on a variety of services rather than just cash benefits, plus discretion to institute family caps and a number of other policy measures without getting a waiver from the federal government. But states also confronted new rules, such as a five-year limit for receipt of benefits under the TANF program. (How states used their discretion will be discussed later in the chapter). Despite the narrowed range of issues, reauthorization of TANF proved contentious—scheduled for 2002, it was not enacted until 2005—as the Bush administration and congressional Democrats squabbled over funding levels and forcing states to meet stricter work requirements. In short, seventy years of experience with AFDC and TANF, suggests that the potential for income-tested cash assistance (other than through the tax mechanism, to be discussed) to serve as a mechanism to build a work-based safety net is extremely low. Time limits on receipt of TANF make this program impractical as a vehicle for long-term assistance, and the public appetite for reversing the TANF time limit appears quite limited.

Tax-Based Transfer Policies Tax-based policies, including the Earned Income Tax Credit, Child Tax Credit, tax benefits for employer-provided health insurance and pensions, and mortgage interest deductibility, constitute what Christopher Howard (1997) has called America’s “hidden welfare state.” Tax-based policies have a major shortcoming in

304

/

The Politics of Low-Income Families

terms of benefiting low-income families, however: if they are based on a deduction to taxable income, they are likely to be of benefit primarily to higher-income families. As Christopher Howard (1997) puts it, “To the extent that tax expenditures are targeted, it is the rich who benefit and not the poor” (34). Even flat-rate tax credits are likely to be major helps only if they are refundable, since lowincome families pay little in income tax. But the refundable portion of tax credits does show up in the federal budget just like other expenditures—that is, these expenditures shed their “cloak of low visibility,” which may make them more susceptible to budget retrenchment initiatives. As Howard has noted, the tax expenditure welfare state also has a politics distinct from programs that are funded through direct expenditures. Initiation of taxbased social programs is much less concentrated in a few “big bangs” than social insurance programs. Tax expenditures are also more likely to feature a prominent role by moderate and conservative policymakers, because they are seen as more “marketlike” and “avoiding government bureaucracy” (Howard 1997, 10–11; see also Howard 2007; Hacker 2002, 2006). These tax expenditures continued to grow in the tough budgetary climate of the 1980s and 1990s, and have proved highly resilient in the face of continued retrenchment pressures. Digging beneath the surface to look at two major tax-based programs targeted at low-income families confirms much of this conventional wisdom, and suggests several further nuances. The origins of the Earned Income Tax Credit, ably told by Howard, clearly fit the conventional wisdom on tax expenditures. The EITC had its origins in Senator Russell Long’s opposition to the income guarantees in Nixon’s Family Assistance Plan, and was intended as a relatively modest work incentive that would compensate low-income workers for rising payroll taxes. After several failed attempts to include the EITC in tax legislation, Long succeeded in 1975, where the EITC provision “was a small part of a larger revenue bill; . . . no hearings were held or votes taken specifically concerning the EITC. . . . It generated little debate and reflected little input from interest groups . . . [and] moderate to conservative members of the revenue committees were instrumental to its passage” (Howard 1997, 72). The EITC also benefited in this period from its ability to serve multiple functions to different people: an income supplement to Democrats, an alternative to minimum-wage increases and child-care subsidies for Republicans, and, perhaps most important, a way to make tax legislation look less regressive in the published tables on distributional effects by income class that now accompany legislative proposals on their journey through Congress (Howard 1997, chapter 7). Initial expansions of the EITC were enacted with bipartisan support. It would be a mistake to overemphasize bipartisan support for the EITC, however. In particular, the huge EITC expansion enacted in 1993 was a core element of Bill Clinton’s efforts to eliminate poverty among low-income working families with a full-time worker. But it was enacted as part of the 1993 Clinton tax bill that did not get any Republican support in Congress. And although the Internal Revenue Service and General Accounting Office began to raise concerns about abuse of the EITC early in President Clinton’s first term, Republicans were rebuffed in their efforts to cut the EITC significantly in the welfare reform packages vetoed by

/

305

Making the Work-Based Safety Net Work Better

the president in 1995. The program has remained essentially untouched by cutbacks since then (Weaver 2000, 241–42, 265). The expansion of the Child Tax Credit both parallels and diverges from that of the Earned Income Tax Credit. The expansion of the EITC was backed by a long line of research and argumentation among economists and groups in both parties who favored increasing work incentives. Creation of the Child Tax Credit in 1997 and its subsequent expansion was supported by a different coalition. It was strongly supported by Christian conservative groups who wanted additional supports for traditional families where one parent stayed home. In its initial form it was modest in size and refundable only for families of three or more children. President George W. Bush pledged to increase the Child Tax Credit as a major element of his 2001 tax bill, but without a further extension of refundability to low-income families with very limited income tax liability. “Inside the Beltway” considerations were critical in the expansion of refundability in 2001. Several moderate Republican senators expressed unhappiness with the distributive skew of the overall tax bill toward upper-income groups, and liberal advocacy groups united around a proposal that was carefully tailored to appeal to Republicans who rejected “handouts” to nonworking families by making eligibility for refunds subject to an earnings requirement roughly equivalent to the income of one full-time minimum-wage worker. This provision won inclusion in the tax-cut bill because moderate Republican senators looking to make the bill less regressive signaled that they would not vote for a tax bill that did not include it (see Rich 2006, 196–201, Howard 2007, 85-87). It has become a major source of income to low- and moderate-income families: in fiscal year 2009, outlays for refundable Child Tax Credit payments are expected to total almost $16.8 billion; this in addition to a tax expenditure (mostly to middle-class families) that will reduce revenues by $30 billion dollars.14 In short, tax-based cash transfers have been a major political success story, with significant bipartisan support (though there has been increased resistance from conservative Republicans). They have become a major source of income transfers to low-income working families in the United States since about 1990, proving to be far more politically acceptable than either the negative income tax proposals of the 1960s and 1970s or payments under AFDC and TANF. By 2005, federal cash payments for the refundable portions of the Earned Income Tax Credit and Child Tax Credit were more than four times cash assistance under TANF. However, they remain seriously problematic in providing a steady income stream for those families, since they are generally only available as once-a-year tax refunds. Moreover, it is not clear how much room there is for further expansion of the EITC, since further permanent increases in its value would either be very extensive as benefits are extended further up the income scale or require steeper phase-out rates, with attendant work disincentives.

Income-Tested In-Kind Assistance: A Mixed Bag We might expect that the story of income-tested in-kind benefits (notably health care and nutrition assistance) for low-income families would be similar to that for 306

/

The Politics of Low-Income Families

cash assistance for that same group. In fact, however, it has been very different, and has also exhibited substantial internal diversity. Spending in this sector is dominated by Medicaid, created in 1965, and its younger sibling, the State Child Health Insurance Program (SCHIP), created in 1997. Both programs share costs with the states, although Medicaid is an entitlement program for many of the individuals it serves and has a less generous matching formula than SCHIP. The story for expenditures in both programs has been one of virtually uninterrupted growth, despite recent efforts to control Medicaid costs. Reflecting the liberalizing policy changes in Medicaid and the creation of SCHIP, spending by the federal government and the states on health care for low-income families doubled in real dollar terms between 1988 and 1996, and it virtually doubled again by 2003.15 The politics, however, is a good deal more complicated than the overall expenditure pattern suggests. Repeated expansions of Medicaid eligibility for children, and the breaking of the link between Medicaid and AFDC-TANF, contributed to increased coverage (see Mann, Rowland, and Garfield 2003). But politicians at both the federal and state levels have also repeatedly tried to bring escalating Medicaid costs under control, including making it easier for states to enroll Medicaid recipients in managed care plans that seek to lower costs by giving providers an incentive to reduce the volume of services provided. In 1995, Republican congressional leaders included major Medicaid and Medicare cuts in their politically popular welfare reform legislation, but President Clinton used defense of the health programs as a justification for vetoing these bills. Republicans also had to take Medicaid cuts out of the welfare legislation they sent to the president in 1996 (Weaver 2000, chapter 12). The efforts at cutting Medicaid that have succeeded have frequently been directed at Medicaid providers rather than directly at cutting eligibility and benefits for recipients, although the cuts almost certainly have indirect effects on accessibility of care. The SCHIP program was initially enacted as a compromise in 1997—a capped grant program to the states rather than an individual entitlement. Despite the political popularity of providing insurance to low-income children, SCHIP has also increasingly become the subject of partisan debates. In the summer of 2007, congressional Democrats sought to more than double program funding, relying heavily on tobacco taxes as a revenue source. Clearly the Democrats’ objective was to frame the issue in a way that would force President Bush and congressional Republicans to choose between backing a threatened presidential veto and appearing to be “antikids” or agreeing to the Democratic spending request that would alienate their conservative basis (as well as the tobacco lobby) by agreeing to a major SCHIP expansion. Republicans and their allies, on the other hand, sought to frame the SCHIP expansion as “Hillarycare on the installment plan,” and argued that it carried the serious moral hazard of encouraging low-wage workers who currently have private health insurance to drop it in favor of SCHIP; opposition was important both philosophically and to shore up a disillusioned Republican base before the 2008 elections (see Bucci and Beach 2007).16 In short, a program born in a spirit of bipartisan compromise was increasingly mired in partisan and ideological wrangling. Efforts by the congressional Democratic leadership to enact a major expansion of SCHIP in 2007 (more than doubling spending /

307

Making the Work-Based Safety Net Work Better

over five years) twice fell victim to vetoes by the president, and Democrats were unable to overturn those vetoes in the House of Representatives (Lee 2008).17 Democrats quickly moved a major expansion of the SCHIP program through Congress in the first month of the Obama administration, with more coverage, and less Republican support, than the 2007 bill. Nutrition programs have had a different set of political dynamics than Medicaid and SCHIP. The major federal nutrition programs—Food Stamps, schoolbased nutrition, and the Special Supplemental Feeding Program for Women, Infants and Children (WIC) programs—have all benefited from broad public concern that vulnerable citizens, especially children, should not go hungry. But they have also relied heavily on coalition politics—assembling legislative packages that can pass among legislators with different preferences. Inclusion in broader agricultural legislation has facilitated vote trading between urban legislators who were in favor of nutrition programs and rural legislators who backed commodity-price supports (Ferejohn 1986). For example, the School Lunch program, which was made a permanent program in 1946, was initially a program that indirectly increased income to farmers and provided modest subsidies and modest benefits, primarily to middle-class families. The program was dramatically expanded during the early 1970s to provide increased aid to low-income families. But along the way, Congress rejected the Johnson administration’s efforts to increase targeting by taking money away from existing middle-class constituents, and never seriously considered proposals to make the program a universal one (Steiner 1976). Creation of the Special Supplemental Feeding Program for Women, Infants and Children (WIC), owes much to a local nutrition program in Memphis that caught the attention of Senator Hubert Humphrey, aided by media coverage “within a political context characterized by a national concern for hunger in America and with civil rights” (Imig 2006, 22). This is not to say that federal nutrition programs have been without controversy. The Food Stamp program has been particularly vulnerable to charges of fraud and abuse, and until the advent of Electronic Benefits Transfer (EBT) technology, Food Stamp transactions were also highly visible to the broader public. But Food Stamp cutbacks after the “Gingrich revolution” of 1994 were heavily concentrated among immigrant non-citizens, and Republican efforts in 1995 to cut school-based nutrition programs were mostly defeated (Weaver 1996). Moreover, after clear evidence emerged during the shift after 1996 to more workfocused policies that people leaving AFDC or TANF cash assistance were also losing Food Stamp and Medicaid benefits to which they remained entitled, concrete steps were taken, with bipartisan support, to rectify the situation.18 Thus incometested health and nutrition programs appear to have at least made a transition among some (though clearly not all) politicians from being perceived as “welfare” to being perceived as “critical work supports”—much firmer ground for continued funding support. What generalizations do these outcomes suggest about the politics of incometested in-kind transfers? Clearly there are important differences across programs, with health-care programs almost certainly showing more potential for growth

308

/

The Politics of Low-Income Families

than nutrition programs. Overall the political prognosis is not nearly as dire as it is for income-tested cash assistance.

CONSTRAINTS ON FUTURE PATHWAYS The argument thus far is that different “streams” of programs aiding low-income families have distinctive political logics, and distinctive outcomes based in part on factors such as whether they are contributory (social insurance) or not, and whether policymaking is annual and highly visible or intermittent and not visible (tax expenditures). But there are also important differences in outcomes within streams of programs and within programs over time. Underlying the differing program outcomes in the last section are a set of underlying causal forces that appeared repeatedly over time and across programs. These forces range from public attitudes toward the poor to the issues of whether the program delivers most of its benefits to a constituency that the public views sympathetically or unsympathetically or whether programs are nationally uniform or subject to variations in state, politics, ideology, and fiscal capacity. In this section I turn from looking backward at the evolution of programs to looking more explicitly at those underlying conditions and what they suggest about future directions for efforts to build a work-based social safety net for low-income families. Although each of these conditions may affect each stage of the policymaking process (and those stages are in fact usually overlapping in time rather than sequential), I discuss them in rough order from those that are particularly important for the way that problems are defined and reach policymakers’ agendas to those that have the greatest consequences for the way that alternatives are formulated and those that affect what policies are adopted and how they are implemented.

Continuation of “The Great Risk Shift” A first factor that is certain to keep policy proposals to help low-income families on the agenda is what Jacob Hacker has labeled “the great risk shift”—increased uncertainty in employment, pensions, and health insurance coverage (Hacker 2006). Health insurance coverage is a particular concern for low-income families: the percentage of families with nonelderly workers with incomes between 100 and 199 percent of the poverty line who are offered health insurance at work dropped from 64 percent in 1998 to 60 percent in 2005 (comparable figures for those at or above 400 percent of the poverty line were 91 percent and 90 percent) (DiJulio and Jacobs 2007). Overall, 40.1 percent of nonelderly low-income (less than 200 percent of the poverty line) adults lacked health insurance in 2005, more than twice the rate (18.9 percent) for children in those families (Kaiser Commission on Medicaid and the Uninsured 2007). Moreover, these figures understate both the growing number of Americans who lack health insurance at some point

/

309

Making the Work-Based Safety Net Work Better

during a year and the drop in employers paying the full cost of coverage (Hacker 2006). Increased anxiety among the public about health insurance and employment and pension security does not mean that politicians will enact legislation that deals with these concerns. But it does mean that the health insurance issue in particular is likely to be on the agenda repeatedly, and that America’s middle class may see the issue as one that concerns them as well, giving advocates for low-income families multiple opportunities for “salami-slicing” incremental policy changes. In theory, concerns about increased employment insecurity could make a similar dynamic of incremental expansion possible for Unemployment Insurance, especially if the current financial crisis leads to a large and lengthy increase in unemployment. However, the relatively low incidence—at least so far—of prolonged unemployment and its concentration among the most politically disengaged and powerless sections of the populace, in combination with a public tendency to view Unemployment Insurance as somewhat akin to “welfare” rather than an earned social insurance right and general federal deference to states in Unemployment Insurance policymaking make a major, permanent expansion of the program less likely.

Public Opinion and Issue Framing A second critical factor in policy toward low-income families is public opinion toward the causes of poverty and toward the programs that seek to eliminate or alleviate poverty. Politicians are more likely to respond to an issue when there is a widespread perception that a problem exists, government (as opposed to other actors) can and should do something about it, and there is something specific that government can do that would be useful.19 There is much continuity in American attitudes concerning policy toward lowincome families.20 In particular, there is an emphasis that the able-bodied should be expected to work (although with a hesitation in applying this to mothers of young children that appears to be declining over time). Another consistent theme in American attitudes is that policy should promote greater equality of opportunity (through education and training) and a minimal safety net rather than equality of outcomes. With regard to specific programs, the public consistently supports additional spending when questions ask them about “the poor” rather than “welfare,” and they offer greater support for in-kind assistance (especially medical care) than cash assistance for the able-bodied poor (Heclo 1986, 326–32; Heclo 1994, 400–05). Martin Gilens (1999) and others have shown that heavily racialized images of the poor—with African Americans frequently identified as a lazy “nondeserving poor”—have long-standing and deep roots. Opposition to “welfare” in particular is related to perceptions that too many recipients—a disproportionate share of whom are racial and ethnic minorities—could get by without welfare if they really tried.21 Contrary to the hopes of welfare reform proponents, enactment of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) in

310

/

The Politics of Low-Income Families

1996 does not appear to have fundamentally shifted attitudes toward low-income Americans and toward programs that serve them. Images of the poor remain heavily racialized (Soss and Schram 2007). The fall in TANF caseloads and increased emphasis on work appear to have led not to increased public support for antipoverty programs, as advocates for the poor had hoped, but to something closer to a public attitude of disengagement or “benign neglect.” Past economic cycles suggest that a prolonged economic downturn in the wake of the 2008 financial system crisis is likely to lead to an increase in the ratio of Americans who in public opinion surveys state that poverty is due to circumstances beyond individuals’ control relative to those who believe that it is due to insufficient effort. For most voters, however, the broader problem of family poverty is likely to remain less salient than aspects of an economic downturn that affect their personal welfare, such as losses of retirement savings, declining real earnings and home prices, and a credit crunch. Moreover, as Doug Imig (2006, 24) has noted, Americans tend to view parental inattention as the most serious problem facing American families, with some skepticism about the role that government, especially traditional income-transfer programs, can play in dealing with the problems of children. In a 1999 survey, only 9 percent said that “increasing government funding for such welfare programs as AFDC and Food Stamps” would be very effective in helping children, compared to 34 percent for “more government funding for childcare and health care programs,” 52 percent for “more involvement for volunteer organizations dedicated to kids, like the Boy Scouts and YMCA,” and 59 percent for more flexible work schedules from employers, and 68 percent for improvements in quality of public schools. In the same survey, 49 percent of the respondents blamed “irresponsible parents who fail to do their job” for the problems faced by kids, compared to only 37 percent who blamed social and economic pressures on families (Public Agenda 1999, 5–8). These attitudes might appear to be fertile ground for what might, for lack of better term, be called “backlash politics,” a general reaction against programs that benefit low-income families, based on a perception that many of those program recipients are undeserving. In fact, however, that reaction appears to be limited primarily to cash benefits, and to a lesser extent Food Stamps. Noncash benefits, especially those that are more effectively targeted at children (child nutrition programs and SCHIP) have been highly resilient in resisting retrenchment and appear likely to remain that way. While the general public remains largely ignorant of the complex ways in which many low-income families combine work and a maze of public benefits, most of the programs that form the basis of the current work-based safety net enjoy substantial, albeit diffuse and uninformed, public support.

The “Politics Of Permanent Austerity” A third major influence on future policies toward low-income families is what Paul Pierson (2001a) has called a “politics of permanent austerity.” Contrasted

/

311

Making the Work-Based Safety Net Work Better

with the earlier postwar period through the early 1970s, which Elliott Brownlee labeled “the era of easy finance”(Brownlee 1996), the period since then has been much less favorable to financing new social policy initiatives. Figure 9.2 provides a very crude indicator of “fiscal slack,” or room for financing new policy initiatives. The line in figure 9.2 shows the total federal surplus or deficit, including trust funds surpluses or deficits, expressed as percentages in GDP. The bars show the “federal funds” deficit for the same fiscal year—the fiscal balance excluding surpluses from federal trust funds. As the figure shows, the federal government has been heavily dependent since the early 1980s on trust fund surpluses to keep federal budget deficits down. Only once in the past fifty years, in fiscal year 2000, has there been a federal funds surplus, and that one was tiny. The year that President Clinton proposed his health-care initiative, 1993, followed one of the worst years for the federal funds deficit in the post-1950 period. Political and policy developments since the late 1970s have exacerbated pressures for austerity in social spending. On the revenue side, the perceived tax revolt symbolized by the passage of Proposition 13 in California made politicians very wary of explicit increases in taxes, while the indexation of federal income tax brackets in 1981 made it much more difficult to generate general revenue increases stealthily through bracket creep. The clear signal sent by Republicans in 1983 that they would not accept any further increases in Social Security payroll tax essentially ended payroll tax increases as a source of higher revenue. On the expenditure side, fiscal pressure was eased through the 1990s by a decline in defense spending as a share of GDP, but this was reversed dramatically after September 11, 2001. A tentative consensus first on a norm of overall budget balance and then on running a balanced budget excluding Social Security account surpluses emerged in the late 1990s, but commitment to fiscal discipline collapsed in the aftermath of September 11. Budget spending caps and “PayGo” rules for mandatory spending were abandoned at the end of the 2002 fiscal year.22 Most congressional Republicans followed the G. W. Bush administration’s lead in abandoning their traditional emphasis on balanced budgets in order to defend and expand on the parties’ 2001 tax cuts while continuing to press for spending cuts.23 The economic downturn that began in 2008 created a consensus on the need for an economic stimulus package. That opportunity was used by the Obama administration and congressional Democrats to introduce a number of measures that will benefit low-income families. Most of these provisions are temporary, however, rather than creating permanent new program commitments. Overall fiscal constraints on programs that help low-income families seem almost certain to deepen in coming years.24 Increased spending on homeland security, defense, and health care (including rising expenditures for the new Medicare prescription drug benefit) and the huge commitments undertaken in the wake of the fall 2008 financial crisis will all add to fiscal stress. A declining ratio of workers to retirees will end Social Security surpluses in another decade, putting an end to the “cash cow” that has helped finance the federal funds deficit for a quarter century. Thus any major new initiatives are likely to meet increased political resis-

312

/

0

5 19

/

54

19

58

19

66

19

70

19

74 19

Year

77 19

Federal Funds Surplus or Deficit

62

19

81 19

85

19

89

19

93

19

97

20 Total Federal Surplus or Deficit

19

Federal Government Budget Surpluses and Deficits, 1950 to 2007

01

20

05

Source: Budget of the United States Government, Fiscal Year 2009. Washington, D.C.: Government Printing Office, Tables 1.2 and 1.4.

−8%

−6%

−4%

−2%

0%

2%

4%

FIGURE 9.2

Percentage of GDP

Making the Work-Based Safety Net Work Better

tance, and they are likely to be subject to increasingly stringent “PayGo” rules at the federal level. Some insight into what is likely to happen in future rounds of fiscally induced belt tightening can be gained by looking at past rounds of budgetary restraint. Table 9.2 shows summaries of percentage contributions to deficit reduction in three deficit-reduction packages, in 1990, 1993, and 1997. They clearly show that the composition of those packages is influenced both by the strength of Republican and Democratic political forces and by the overall fiscal situation. The 1990 package shows the greatest balance between mandatory spending changes, changes in revenues, and changes in discretionary spending, although changes in mandatory spending played the smallest role. The 1993 package, with Democrats in control of both Congress and the presidency, not surprisingly has the greatest reliance on tax increases. And the 1997 package, with a much improved fiscal outlook as well as Republican majorities in Congress and a weakened Democratic president, features major tax cuts along with cuts in both entitlements and discretionary programs—and the smallest overall deficit-reduction package. The lesson for the future, in short, is that when a fiscal crunch hits, the relative balance of partisan forces in the Congress and the executive will make a big difference in how the pain of retrenchment is distributed. This may well be the most important consequence of Democratic gains in the 2006 and 2008 elections for policies toward low-income families—assuming that those gains are sustained over time. Even if the recent trend toward declining fiscal discipline persists, however, Democrats remain electorally vulnerable to charges that they are the party of “tax and spend.” They are, therefore, in a poor political position to propose major new social policy initiatives targeted on families with very low incomes. Policymakers are likely to be consumed with the questions of how to provide President Obama’s pledge of a “middle-class tax cut” while providing a short-term fiscal stimulus and later filling the fiscal hole created by the Bush tax cuts and the 2008 economic financial sector meltdown. Major new revenue-raising measures—a broad-based national consumption tax or carbon tax—are likely to remain political nonstarters. In theory, money could be freed up for spending on the poor by cutting other programs. However, existing spending commitments, notably the two major programs for the elderly, Social Security and Medicare, are backed by powerful, organized constituencies and strong legal entitlement as well as felt (by the constituencies and others) entitlement. These factors make substantial retrenchment in these programs—and especially cutbacks that are framed by advocates of aid for low-income families as necessary to free resources to help those families— very unlikely. Many spending programs for low-income families face additional procedural obstacles. Several programs for low-income families that enjoy the privileged status of entitlements or tax expenditures, notably the TANF block grant and the refundable Child Tax Credit, are not indexed for inflation (Steuerle 2003). Much spending on services for children is discretionary spending that must go through the appropriations process each year.

314

/

The Politics of Low-Income Families TABLE 9.2

/ Distribution of Deficit Reduction Across Program Categories in Recent DeficitReduction Packages

Percentage of Deficit Reduction

Legislation Omnibus Budget Reconciliation Act of 1990 Omnibus Budget Reconciliation Act of 1993 Balanced Budget and Taxpayer Relief Act of 1997

Revenue Changes

Mandatory Discretionary Debt Five-Year Spending Spending Service Totals (Billions Changes Changes Changes of Current $$)

32.8

15.6

39.4

12.2

–482

55.7

17.8

15.9

10.9

–433

–67.8

90.7

75.4

1.7

–118

Source: Robert Keith, Deficit Impact of Reconciliation Legislation Enacted in 1990, 1993 and 1997, Congressional Research Service Report for Congress, Order Code RS22098 (Washington, D.C.: updated August 30, 2006).

These constraints may tempt politicians to “legislate by regulation” to keep direct budget costs down, using mechanisms such as minimum-wage increases and health-care coverage mandates, but this is likely to encounter strong resistance from business, which will be discussed later in this chapter. Efforts to use the tax system for social policy initiatives will also hold substantial appeal for many policymakers, although room for further permanent expansions of the Earned Income Tax Credit appears to be limited (Pierson 2001b, 73).

Institutional Polarization and Institutional Fragmentation A fourth factor that will shape policies toward low-income families is the partisan composition of government and institutional fragmentation within national governing institutions. Figure 9.3 provides a very rough indicator of changes over time in the balance of political forces in national legislating institutions—the two houses of Congress and the presidency. The indicator (described in notes to the table) combines nine measures of party institutional control as well as (for both chambers of Congress) ideological shifts on a left-to-right scale. The indicator is very crude, but it does provide a visual reminder of just how unusual the institutional configurations of the New Deal (especially 1935 to 1938) and the Great Society years (especially 1965 to 1966) were, and thus how unlikely it is that the torrents of social legislation—and especially very expensive social insurance programs—that those periods produced will be repeated. Each of the peaks of Democratic institutional power since the Great Society (during the Carter years and the first two years of the Clinton administration) has been at levels significantly lower than those during the Great Society. Indeed, if we project that the new 111th

/

315

−15.00

−10.00

−5.00

0.00

5.00

10.00

15.00

/

Liberal Control of Federal Executive and Congress, 1929 to 2008

Year

29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99 01 03 05 07 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20

FIGURE 9.3

Liberal Control of Institutions Score

Sources: Congressional seat share data is from Norman J. Ornstein, Thomas E. Mann, and Michael J. Malbin, Vital Statistics on Congress, 2008, Washington, D.C.: Brookings Institution Press, 2008, pp. 46–53. DW-NOMINATE first dimension scores are from “Party Medians from DW-NOMINATE Congresses 1-110, available at: http://voteview.ucsd.edu/pmediant.htm (accessed January 28, 2009). Scores in figure 9.3 are the additive results of the following components: a. A measure of party control of the presidency, with a Democratic president getting a score of +1, and a Republican president a score of −1. b. A continuous scale of the Democratic presidential candidate’s share of the presidential popular vote, with a Democratic share of 60 percent counting as +1, 50 percent as 0, and 40 percent as −1 (thus scores of greater than +1 and less than −1 are possible). c. A continuous scale of the Democratic presidential candidate’s share of the electoral vote, with a Democratic share of 90 percent counting as +1, 50 percent as 0, and 10 percent as −1 (thus scores of greater than +1 and less than −1 are possible). d. A measure of party control of the U.S. Senate, with Democratic control getting a score of +1, and Republican control a score of −1. e. A continuous scale of Democratic seat share in the U.S. Senate, with a Democratic share of 60 percent counting as +1, 50 percent as 0, and 40 percent as −1 (thus scores of greater than +1 and less than −1 are possible). f. Standardized DW-Nominate first dimension chamber median scores for the entire Senate for Congress, with scores of 1.5 standard deviations from the mean equaling a difference of 1 on the scale. g. A measure of party control of the U.S. House of Representatives, with Democratic control getting a score of +1, and Republican control a score of −1. h. A continuous scale of Democratic seat share in the U.S. House of Representatives, with a Democratic share of 60 percent counting as +1, 50 percent as 0, and 40 percent as –1 (thus scores of greater than +1 and less than −1 are possible). i. Standardized DW-Nominate first dimension chamber median scores for the entire U.S. House of Representative for Congress,

Making the Work-Based Safety Net Work Better

Congress (2009–2010) will have the same House and Senate ideological means as the very liberal 89th Congress (1965–1966), the overall “liberal control” index in 2009–2010 is likely to be lower than in the first two years of the Carter administration. In 2008—a very bad year for Republicans—144 Republicans elected to the House were elected with 55 percent of the vote or more, a margin generally considered safe from all but the largest swings in partisan support among the electorate. With the possible exception of a move toward more comprehensive healthcare provisions, Democratic political gains in the 2008 election are unlikely to provide the commanding majorities in Congress needed for a major expansion of social insurance programs or another “paradigm-shifting” change in programs benefiting low-income Americans in the near future. Not directly shown in figure 9.3 but also relevant for social legislation is the decline in ideologically moderate legislators in Congress that has been documented by Nolan McCarty, Keith Poole, and Howard Rosenthal (2006) and others. The shrinking of the pool of moderates makes it difficult to fashion broad compromises that will win overwhelming support in Congress.25 Moreover, as Jacob Hacker and Paul Pierson (2005, 53–55) have noted, many congressional Republicans face pressures not to depart from the positions of activist groups, such as the anti-tax Americans for Tax Reform, by deviating from the party line on issues that matter to the party’s conservative base. The institutional position of Democrats is particularly constraining when combined with the decline in fiscal slack noted earlier. Even limited forms of bipartisan cooperation, such as the brief political era of bipartisan support for expansion of the Earned Income Tax Credit beginning in the early 1980s or the much more tenuous bipartisan agreement (or more accurately, Republican prodding and a Democratic split leading to the latter’s acquiescence) that led to the 1996 welfare legislation is likely to be intermittent at best. Republicans have invested too much in portraying themselves as the party that is tougher on welfare cheating, and the bulwark of resistance to the “tax and spend” and “spread the wealth” Democrats, to abandon those positions. Indeed, as the recent history of the SCHIP program suggests, the trend appears to be in the opposite direction. Given the continued closeness of political competition between the parties and the search on both sides for “wedge” issues that will divide and create negative public images of the other party, extensive cooperation across party lines is likely to be limited and “moment-specific,” such as the expansion of the Child Tax Credit enacted in 2001.

Federalism Federalism remains a fundamental fact of American political life and American social policy. The literature on the impact of federalism on policy suggests at least four distinct effects on policy toward low-income families. First and most optimistically, it may provide an opportunity to experiment with ideas that may not yet have broad national acceptance (called the “laboratory of democracy”). As evidence from state-level experiments emerges, “good” or “best” practices may be

318

/

The Politics of Low-Income Families

diffused across the states or adopted as national policy by the federal government (Karch 2007). But federalism may also have other, less benign, effects. It may lead to a mutual buck-passing by both levels of government—and thus governmental inaction—in policy sectors where there is concentrated opposition to government action from some societal interests (Harrison 1996). A third possible effect is variation across states based on the characteristics of those states, including fiscal capacity (notably per capita income) and state political environments (political culture, and ideology of the broad public and elites, as well as interest group pressure, partisan balance in the legislature, and so forth). Fourth, federalism may lead to a “race to the bottom,” especially in programs such as the Temporary Assistance to Needy Families program and its predecessor, AFDC, as well as Unemployment Insurance and General Assistance, where states have discretion (or complete control) over program eligibility, benefits, and associated program features. Because states have an interest in keeping their tax rates down, states might provide lower levels of transfers and services than they would otherwise prefer out of a fear that their policy efforts may encourage in-migration (and discourage out-migration) by persons who are likely to be high consumers of transfers and other government services if they are more generous than neighboring political jurisdictions (see, for example, Berry, Fording, and Hanson 2003). Moreover, the “welfare migration effect” need not be real for the effects of competitive federalism to cause policy to become more stingy or strict: it is enough that policymakers fear that it is real, or fear that they will be blamed for it by voters, even if they do not themselves believe in the welfare migration effect (Schram and Soss 1998; Peterson and Rom 1990). Which one of the several potential effects of federalism (laboratories of democracy, race to the bottom, etc.) is dominant? Available evidence suggests major differences across programs. Moreover, a race to the bottom and separating out state competition effects from other potential causes of interstate policy variation (for example, differences in state public opinion or fiscal capacity) turns out to be methodologically very complicated, and studies attempting to find it have produced mixed results. There does appear to be evidence of interstate competition in the Unemployment Insurance system, especially with respect to effective tax rates.26 The lack of change in UI’s federal taxable wage base for more than twenty years suggests that mutual buck-passing may be at work as well (see Kletzer and Rosen 2006). The effects of federalism on the Temporary Assistance to Needy Families program and its predecessor, AFDC, suggest an even more complex combination of federalism’s potential effects. Most studies find a very modest migration pattern consistent with the “welfare magnet” hypothesis.27 Studies of welfare policymaking suggest that states allowed their benefits to erode over time with inflation under the old AFDC program, although actual cuts in the nominal level of AFDC benefits were infrequent (see, for example, Volden 2002). Tight federal restrictions on most program dimensions in the AFDC program meant that benefit levels were the major type of program variance.28 Higher benefits in a state were generally associated with greater fiscal capacity, a lower African American share of the

/

319

Making the Work-Based Safety Net Work Better

AFDC caseload, and more liberal overall public opinion, with mixed results on state “neighborhood” effects (see, for example, Wright 1977; Tweedie 1994; Peterson, Rom, and Scheve 1998). In the early 1990s, states were allowed to apply for waivers to obtain increased discretion over a number of AFDC program elements, such as increased earnings disregards and family caps to discourage further nonmarital births to welfare recipients. The major effect of federalism on AFDC waivers seems to be less a race to the bottom than simply increased variation across states based on differences in state political and demographic characteristics and fiscal capacity. With regard to punitive, or “stick,” policies, including more stringent work and pregnancydiscouraging policies such as family caps, a number of studies have found that these policies tend to be enacted in states with higher African American caseload shares (see in particular Schram et al. 2001; Fellowes and Rowe 2004). State choices of “carrot” policies (such as more generous earnings disregards for working parents and more generous spending on child care) are generally associated with states that have greater fiscal capacity and higher TANF block grants per low-income child (Gais and Weaver 2002). Data on state practices using the discretion granted by TANF after it replaced AFDC in 1996 suggest that states are not engaged in widespread emulation of “best practices” copied from other states, although there has been some elimination of “worst practices”—for example, severe restrictions on auto assets for TANF recipients who are now expected to work and need reliable transportation. In the past few years, states have slowed experimentation in TANF, and seem to have settled on a set of policies that reflect their economic, social, and political characteristics rather than either emulation of best practice or a race to the bottom. A recent study of three work-support policies—state EITCs, extension of Medicaid coverage, and state minimum wage laws—in the states by Nicole Kazee (2007) also finds complex patterns in state policy choices for programs targeted at low-income families. She found that roughly half of the states “follow consistent patterns and can be considered either High Work Support or Low Work Support states,” with the “former generally located in the Northeast, politically liberal, and wealthy,” whereas the latter are “Southern and Western, politically conservative, and generally poorer”(29). Generous Medicaid policies and higher state minimum-wage laws were particularly likely to be affected by partisan variables. Although a number of states have incrementally expanded health-care access for low-income families, state movement toward universal health-care access has been politically much harder to achieve. Lawrence Brown and Michael Sparer (Brown and Sparer 2001; Sparer 2004), in a review of state policy initiatives to move toward universal health-care coverage in the 1990s, found that those efforts tended to be stalled both by opposition from business and insurers and by regulatory obstacles in the form of federal ERISA (Employee Retirement Income Security Act) preemption. In recent years there has been a revival in state-level efforts to achieve universal or near-universal health insurance coverage through varying combinations of individual (rather than employer) mandates, government subsidies for low-income residents, requirements that employers spend a minimum percentage of payroll on health insurance or face financial penalties, and levies on 320

/

The Politics of Low-Income Families

health-care providers. But despite an increase in legislative activity—legislatures or legislative committees in ten states considered universal care legislation in 2006 (National Conference of State Legislatures n.d.)—enactment has as of early 2009 made little progress outside the New England region. Clearly federalism can have several distinct effects on policies affecting lowincome families, depending on the specific attributes of a policy sector and program. “Laboratory of democracy”-type effects of policy emulation are most likely to occur under the following conditions: •

Channels for transmitting policy lessons between states are strong.

•

An innovation saves money for states adopting it or costs them little.

• An innovation has both policy and political advantages that are clear to policymakers. •

Political opposition is limited or politically marginal.

•

Barriers in federal law are minimal.

Where these conditions are absent, progress is likely to be slower, smaller, and possibly nonexistent. In the case of universal health care, for example, it seems likely that if the federal government does not enact comprehensive health-care reform in the near future, at least a few states will adopt near-universal health-care coverage, and probably several more will adopt legislation that achieves nearuniversal coverage for children. But even if this occurs, it does not mean that diffusion across states would be quick or universal in the absence of a federal mandate and substantial federal financial assistance. A more likely outcome in the medium term would be something closer to the outcome with TANF “carrot”focused work promotion policies: substantial differences in coverage across states, with states that are richer and more liberal adopting more generous policies. Competitive federalism in the form of buck-passing or a “gentle slide” or even “race to the bottom,” on the other hand, is most likely to occur under the following circumstances: •

A program initiative is costly to state governments.

• The program redistributes money to the poor and the program or its clientele are unpopular (as with AFDC). • The federal government fails to provide strong financial incentives to the states that would make the program more generous or equalize fiscal capacity. • In addition to the previous circumstance, there are direct tax-rate implications that cause business to be attentive and have a stake in lower rates, as in the case of Unemployment Insurance. Both emulation and competitive federalism are limited in their impacts, however: they coexist with differences arising from both state fiscal capacity and state ideological and racial attitudes. Differences based on state fiscal capacity are most /

321

Making the Work-Based Safety Net Work Better

likely to be relevant where a program or policy is expensive to implement and a program’s funding formula requires substantial matching expenditures by state governments and is not highly redistributive to poorer states. Differences based on ideology and racial attitudes are most likely to occur in programs that are perceived by the public to tolerate or reward not working and that have a racialized image.

Lack of Engagement by the Business Community and Organizational Weakness on the Left Another serious obstacle to improving the work-based safety net for lowincome families is the disengagement or hostility of the business community to many of the safety net’s key components. Large employers are preoccupied with issues such as constraining pension and health-care costs more than on the economic position of low-income families. Powerful small-business interests see proposals for universal or quasi-universal health insurance or an expanded unemployment insurance system as deeply threatening to their cost structure. Business in the United States does not speak with a single voice, given the multiplicity of business groups, such as the Chamber of Commerce, Business Roundtable, National Association of Manufacturers, and National Federation of Independent Business, which operate across economic sectors. As a result, individual industries such as health insurers are able to play a role that reflects their disproportionate stake (and fear of losses) in debates on individual programs. Perhaps even more important, small-business interests, through their lobbying groups, notably the National Federation of Independent Business, play an extremely large role in social policymaking wherever a program has the potential to impose costs on business.29 There is substantial support in the business community for income-support programs for low-income workers (notably EITC and Food Stamps) as well as for health-care programs such as Medicaid and SCHIP that make it easier to attract and retain low-wage workers and reduce pressures for higher wages. But these programs are unlikely to be a high priority for business, and to the extent that initiatives are seen as increasing business costs (an employer mandate for health-care costs, for example), they are likely to spark business opposition. Political organizations on the left, meanwhile, offer a weak counterpoint to the political power of business in enacting policies for a work-oriented approach to the problems of low-income families. Labor unions, which acted as powerful advocates for expansion of social insurance programs in the United States, now organize just one-eighth of American workers, with a lower percentage in the private sector than in the public sector. Moreover, many of the key liberal interest groups that form key constituencies of the Democratic Party, such as environmentalists, middle-class women’s groups concerned with reproductive rights, and gay rights groups do not have policies directed at low-income families as central elements of their agenda.

322

/

The Politics of Low-Income Families

Elite Dissensus on Addressing Poverty A final constraint on achieving more than modest expansion of safety-net policies for low-income families in the medium term is what can be termed elite dissensus on the causes of poverty and consequent lack of consensus on what to do about it. More boldly, this can be labeled a lack of plausible intellectual panaceas. Politicians are most likely to spend political capital and propose spending lots of government dollars when doing so is perceived as likely to solve major problems— which may in turn be useful in winning them more votes. In the late 1980s through the early 1990s, several intellectual paradigms vied as alternative solutions to addressing the problems of the poor. For the sake of brevity, these can be summarized as a “Make Work Pay” emphasis on broad universal programs, improved work supports and increased work expectations; a “Disincentive” emphasis on making out-of-wedlock childbearing less financially attractive; and a “New Paternalist” emphasis on more stringent and strongly enforced work requirements. All three of these approaches made strong claims about both the social and political consequences of adopting their particular approach. In so doing they motivated politicians to make the political investment needed to overcome the policy inertia inherent in American political institutions. But the complex of policies adopted by Washington and subnational governments in the 1990s did not make a clear choice between these approaches; instead they borrowed from all three, with significant variation across states (Weaver 2000; Haskins 2006). Studies of families leaving welfare have prompted a more sophisticated understanding among policymakers (if not the general public) of both the ways low-income people combine work and income supports and the barriers to self-sufficiency in the low-wage labor market. These studies have also undermined the credibility of simple “solutions” to the problems fared by low-income families. At present there is no sign of a bold new intellectual paradigm that could spark a major expansion of efforts to aid low-income families. On the other hand, there is little sign of a new paradigm that is likely to strengthen the backlash against low-income families, either. Today most mainstream policy researchers on poverty issues emphasize the multiple and reinforcing sources of family poverty, the need to wage a multifront attack on poverty problems, and the incremental nature of progress that is likely to be made. This complex assessment is an improvement in analysis, but it is not likely to spark bold policy commitments and major investments of political capital by politicians.

STRATEGIC IMPLICATIONS OF ANALYSIS This analysis of the political dynamics of various program streams and underlying policymaking constraints suggests that the opportunities for major permanent extensions of a work-focused social safety net for low-income families are likely to be limited in coming years. This is not to say that there is little hope for a more

/

323

Making the Work-Based Safety Net Work Better

generous and more effective safety net, but rather that proponents of such initiatives will need to choose their battles carefully and devote much attention to social and political coalition-building if they are to be successful. They will probably also need to be more incremental in their proposals than they would prefer. The analysis also raises some important strategic questions: Which policy streams, and which programs within those streams, offer the most opportunities for expansion? What sorts of coalitions should advocates for low-income families try to build, and what strategies should they employ? And what institutional venues should advocates for increased help for low-income families use—electoral politics or courts? National or state politics? In thinking about these questions, I will focus on a few “macro” strategic issues.

Strategic Issue 1: Issue Framing Is Critical, but Not Controllable How a policy proposal is framed, and in particular how it seems to fit with previously held values and perceptions of how the world works, can have an important impact on whether individuals will support that policy or acquiesce in it. Moreover, research on public opinion suggests that “individuals with strong prior beliefs on a topic . . . tend to seize on consistent information with little scrutiny while subjecting challenging information to withering skepticism in ways that allow them to maintain or bolster their prior attitudes”(Peffley and Hurwitz 2007, 998). Given the ambivalence on the part of the public regarding support for aid to low-income families, advocates need to frame their proposals as facilitating work among those who are able to do so and to allow them to move toward self-sufficiency, for example, by breaking down barriers to work.30 Several variations on this broad theme are possible. A universally oriented variant could stress the theme of economic insecurity, arguing that increased uncertainty in employment, pension prospects, and health insurance all call for a broader and sturdier safety net for all Americans, or at least those who meet some set of work expectations. This broader framing, with its focus on middle-class families terrified by fears of lost homes, retirement savings, jobs, and health insurance, has been the dominant theme of the Obama administration’s initial efforts. But it should be recognized that the issues that many low-income families face, such as low skills, unsafe neighborhoods, weak labor-market connections, and nonmarital births, are different at least in their prevalence from the economic security concerns of the “worried middle class” in the United States. An issue-framing and policy agenda that focused primarily on middle-class economic concerns, such as enhanced subsidies for private health insurance and transitional coverage during moderate-length spells of unemployment, would do little to solve the most pressing concerns of many low-income families. A second issue-framing variant would set more explicit targets, similar to those suggested by Bill Clinton in 1992: that if a parent in a family is working full time, that family should not be in poverty and lack health care, but should rather be

324

/

The Politics of Low-Income Families

able to access a package of benefits that will bring them above the (admittedly very low) U.S. poverty line. This framing has the advantage of appealing to American attitudes favoring work, though not to the personal interests of middle-class Americans. But given the serious barriers to full-time work faced by many lowincome (especially single-parent) families, particularly in a severe and prolonged economic downturn, advocates of a work-based antipoverty strategy will need to consider how to develop and build support for effective work-enhancing programs under such an issue framing. Two more promising ways to frame the policy initiatives for low-income families involve improvements in the human capital and productive capacity of young children and adults—what Jane Jenson has called “investing in children” and “social investment” (Jenson 2004; Jenson and Saint-Martin 2003). A recent Brookings Institution volume, One Percent for the Kids (Sawhill 2003), drawing on the Blair government’s initiative to spend an additional 1 percent of GDP in reducing child poverty, is an example of such a framing. The central idea in the social-investment and investing-in-children framings is that investments in individuals, especially those that start off from a disadvantaged background, will pay off for those individuals, for their families, and for overall societal productive capacity. Programs ranging from early-childhood education to job training might benefit from such these framings. They not only appeal to American values of self-improvement and work but also have the additional advantage of potentially engaging business support. As with the second issue framing, however, stretching such an issue framing to give adequate attention to the particular problems of the “disconnected mothers” analyzed by Rebecca Blank (see chapter 7, this volume) would be difficult. Advocates for low-income Americans also need to think through how they can broaden a work-based approach to facilitate inclusion of those who have been largely excluded from past initiatives, notably noncustodial fathers who do not live with their children but who should be helping to support them. An issue framing that emphasizes this clientele is also likely to be more politically problematic, however. For example, one can imagine an issue framing that focuses on the interconnected issues of fatherhood, the problems created by mass incarceration, and the need to create greater connections between young disadvantaged males and their families, communities, and the labor market. But fear of crime among middle-class voters in the United States is sufficiently strong that an issue framing that highlights criminal behavior by a key target of policy is not likely to be very successful in building support for that package of policies. Issue framings that highlight racial disparities in outcomes, either in terms of incarceration or family outcomes, similarly are likely to be unsuccessful because they tend to “prime” attitudes among some whites that hold African Americans liable for those negative outcomes.31 Whatever the issue-framing strategy chosen, it is also critical that advocates not believe that they can control how an issue and reform proposals will be perceived by the public. Lawrence Jacobs and Robert Shapiro (2000) have argued that politicians use polls and focus groups to engage in what they call “crafted talk,” fram-

/

325

Making the Work-Based Safety Net Work Better

ing policy debates in terms consistent with their own policy preferences; advocates attempt to do the same thing. But the debate in 1993 and 1994 over health care and the 1993-to-1996 debate over welfare reform both suggest important limitations on “crafted talk.” Advocates and politicians may be misled by polling data and focus groups to overestimate the political prospects for a proposal if those data suggest that a reform proposal is popular when it is framed in a particular way. The problem is that political opponents may succeed in framing that proposal in another way that makes it much less popular (the Health Insurance Association of America’s “Harry and Louise” advertisements on health-care reform are a classic example of such effective pushback) or to come up with an alternative proposal that is equally popular or even more popular.

Strategic Issue 2: Tsunamis Are Rare Events in Politics, but That Doesn’t Mean You Shouldn’t Prepare for Them One potential pathway for improving the lot and life prospects of low-income families is the most straightforward: a broad set of “pro-poor” policies will be enacted and implemented with the explicit objective of improving economic prospects of the poor through a combination of government-financed transfers and services. There is substantial disagreement, especially on the political left, about what conditions are likely to give rise to a “pro-poor politics,” with some scholars stressing the role of civil disorder in prompting policy response and other scholars seeing little evidence of such a linkage.32 In any case, there is ample evidence that conditions for a broad pro-poor politics do not arise very often. For an explicitly pro-poor policy agenda to be enacted, there must be a fairly widespread sympathy for the condition of the poor, poverty must be highly salient to a very broad segment of the public, the political allies of the poor must have relatively unrestricted power in a political system that is normally riddled with veto points, and they must be able to agree on a set of policies to address those problems. None of this seems very likely—big waves like 1935 and 1965 are rare events indeed, and they are getting much less likely over time. The early exit of John Edwards from the 2008 Democratic presidential race does not inspire confidence that a political campaign focused on income inequality and improving the lot of the poorest Americans will find a broad political following; the focus of both Barack Obama and Hillary Clinton on the middle class proved much more attractive to voters. Thus, one important strategic implication of this analysis is that advocates for low-income families should not count on a big political “wave” or “window of opportunity” that will allow a major initiative such as comprehensive national health insurance to win legislative passage. It should also be remembered, however, that big political waves do sometimes happen in politics, and advocates for low-income families should be prepared with concrete proposals, issue-framing rationales for their approaches, and strategies to win their adoption if such a wave does hit. Just how big and prolonged the tsunami of the

326

/

The Politics of Low-Income Families

post-2008 economic downturn will be remains unclear, but the evidence thus far suggests that initiatives that “incorporate the poor through the political backdoor” (Heclo 1986, 315) are once again most likely to enjoy political success.

Strategic Issue 3: Overreaching and “Strategic Disagreement” Pose High Risks The political scientist John Gilmour (1995) has argued that parties in Congress frequently engage in “strategic disagreement”—failure to agree to a legislative package that is passable in order to get political rewards from blaming the other side for being political obstructionists; reap political rewards from maintaining their distinctiveness from the other party on issues where they have a political advantage with the electorate; and get a more favorable policy package after expected gains in the next election rather than taking a deal now that may make postelection legislative action less likely.33 The analysis presented here suggests that strategic disagreement is usually a poor strategic choice for aiding low-income families. Overreaching and refusal to compromise, followed by unexpected political setbacks, risk squandering real if modest opportunities for policy gains. The strategy pursued by “inside the Beltway” players such as the Center on Budget and Policy Priorities and the Center on Law and Social Policy—organizations that have the confidence of party leaders and experts on poverty issues as providers of useful information and pragmatists who are willing to compromise and will not criticize their political allies as “sellouts” for accepting a deal that may be less than ideal but is the best that is politically achievable—is likely to produce steadier gains for low-income families than more ideologically oriented and intransigent approaches.

Strategic Issue 4: The Camel’s Nose and Salami Slicing A history of programs that aid low-income families suggests that many of them— not just social insurance programs—had relatively modest beginnings that were expanded on over time. The Earned Income Tax Credit and the Child Tax Credit are perhaps the quintessential examples of this, and the SCHIP program may well turn out to be another. Both Medicaid and Food Stamps also were significantly expanded in the clienteles that they served over time. Getting a new program created (the camel’s nose inside the tent) provides an opportunity for incremental expansion in later years. It may build a constituency among those who benefit directly from the program (its clientele), those who administer it and gain resources from it (such as state human resources agencies, for shared-cost programs), members of Congress who played a role in its creation and want to claim credit for its success, and those who evaluate it. In terms of Wilbur Cohen’s salami-slicing strategy, incremental strategies can help at least build a more ade-

/

327

Making the Work-Based Safety Net Work Better

quate sandwich of programs for low-income families, if not a particularly rich one.

Strategic Issue 5: Be Prepared and Be Opportunistic In general, policymaking in the United States is maddeningly slow, but this is not always the case. On occasion, policy packages—including big omnibus packages like those of 1981, 1983, 1990, 1997, and 2001, the financial rescue package in September–October 2008, and the Obama administration’s financial stimulus package in January–February 2009—can move through Congress quite quickly. Advocates for low-income families must anticipate potential “windows of opportunity,” be prepared with a set of proposals that are consistent with the overall philosophy of those packages and their sponsors, and work with their political allies to enact them. The Bush tax cut bill of 2001, for example, was not an obvious target of opportunity for increasing funding for low-income families, but it turned out to be a very important one (through the refundable Child Tax Credit) in a bill that gave most of its benefits to those who are much better off. Other opportunities are likely to arise. For example, Congress will almost certainly be considering a Social Security reform package within the next five years or so, one that is likely to include some significant regressive elements for low-income families. Advocates must be prepared to document those potential impacts, to propose offsetting measures, and to frame them in ways that will appeal to politicians and the electorate. Another potential set of opportunities for policy change has its locus at the state level. The potential for independent state action should not be exaggerated, given both direct fiscal constraints and the potential for interstate competition. But devolution of authority to the states does offer aggressive state officials, especially politically ambitious governors, significant opportunities for policy innovation. Wealthier and more liberal states may be able to use devolution to attempt innovations, especially in the health-care sector, that can offer substantial benefits to low-income families. In addition, organizations of state officials now offer efficient channels for transmitting policy ideas and evaluations of innovations that have been tried in one or a few states to a broader set of potential policy adopters. In short, excessive focus of advocacy efforts at the federal level would be a mistake.

Strategic Issue 6: Some Programs Offer More Opportunities for Growth Than Others Advocates for low-income families should devote much of their limited resources to programs where the growth potential is real. For example, increasing the real size of the refundable Child Tax Credit and indexing it for inflation is probably a winnable battle, whereas others might not be. This is not to say that programs that have limited growth potential should sim-

328

/

The Politics of Low-Income Families

ply be abandoned. The TANF funding stream remains critical in supporting lowincome families and helping them move toward self-sufficiency. Modest increases in the overall size of the TANF funding stream, along with improvements in the formula to give more money to poorer states and an indexing of the total block grant are about the most that can be hoped for in this program, even under unified Democratic control of government. Opportunities for increased funding for many of the purposes served by TANF such as job training and child care may be greater under other existing, or new, funding streams. Of course, states dislike multiple funding streams because of their varied reporting requirements and rules. But they may have to choose between more complexity and more money. Moreover, multiple funding streams, especially if there is some capacity to move money between those streams, can offer states additional flexibility to deal both with budget reductions and with the fluctuating political fortunes of individual programs. Similarly, advocates for low-income families may have to choose between program streams that they think offer the best approaches to dealing with the problems of the poor and those that offer the best political prospects for growth.

CONCLUSIONS This essay has attempted to paint a broad portrait of the constraints that have shaped policies toward low-income families in recent decades and are likely to continue to shape them in coming years. For those who favor a more active government role and bold steps to improving the economic status and prospects of these families, the picture painted here may seem pessimistic, perhaps even unrealistically so. As Joe Soss has pointed out, many of the policy gains made by political conservatives over the past thirty years “seemed unimaginable a short time earlier. They accomplished their goals partly by aiming high and holding tight to their vision and partly by incremental steps that seemed likely to move the broader agenda forward.”34 Advocates for an agenda emphasizing moving lowincome families toward greater self-sufficiency would be well-advised, similarly, to keep in mind both incremental policy objectives and a broader policy vision to help low-income families—and also the relationship between the two. In the short term, much will depend on the strategic choices made by the Obama administration, its political skill in moving its priorities forward, and the responses of congressional Democrats and Republicans. In both the shorter and longer terms, a combination of realism about constraints on action and idealism about what might be achieved are both essential to effective political action.

The author would like to thank Debra Hevenstone for research assistance, and Ron Haskins, Bruno Palier, and especially Joe Soss and the editors of this volume for ex-

/

329

Making the Work-Based Safety Net Work Better tremely detailed and thoughtful comments on earlier drafts of this chapter. William Galston, Jonathan Rauch, Allen Schick, and Isabel Sawhill at the Brookings Institution also provided very helpful comments.

NOTES 1. These are not, of course, the only approaches to promoting greater family selfsufficiency. I do not, for example, give significant attention to efforts to directly alter family structure and reduce the number of nonmarital births, which are a high priority for some analysts and advocates. 2. John Kingdon (1995) similarly refers to “windows of opportunity” that may open briefly at one point in time and close at others. 3. In addition to tight labor-market and labor-market adjustment policies, Wilson stresses child support assurance, family allowances, and a child-care strategy. 4. Social Security also provides benefits to a small number of children of retired workers and to adults disabled before age twenty-two, when the parent is retired, disabled, or deceased. In December 2007, benefits were received by 1.59 million children of disabled workers, 1.37 million children of deceased workers, 298,000 children of retired workers, and 795,000 disabled adult children of retired workers (see Lavery and Reno 2008, 2). 5. In 2002, Social Security filled 21 percent of the child poverty gap, edging out the Earned Income Tax Credit (20 percent of the gap), and far outdistancing incometested programs, including Food Stamps (15 percent), housing assistance (10 percent), SSI and TANF (9 percent), and another social insurance program, Unemployment Insurance (6 percent) (Sherman 2005). 6. In December 2005, for example, the average monthly benefit for a black child OASDI recipient was $385.90, whereas that for a white child recipient was $473.50, reflecting racial disparities in parental earnings (Social Security Bulletin Annual Statistical Supplement, 2006, Table 5.A1.4). 7. For example, extension of Medicare coverage to the disabled and to persons with endstage renal disease (both categories were added to Medicare in 1972) provides substantial assistance to families who are likely to be at high risk of having low incomes, but it has not led to the inclusion of other family members besides the person who is disabled, and these coverage categories are very narrowly framed. The major exception to this pattern is the 1988 passage and quick repeal of Medicare catastrophic coverage. On both the general pattern and the 1988 exception, see Jonathan Oberlander (2003, chapter 3). The services provided under Medicare have also remained relatively fixed rather than evolving into a comprehensive package including long-term care and (until recently) outpatient pharmaceutical coverage. On the prescription drug benefits, see Thomas R. Oliver, Philip R. Lee, and Helene L. Lipton (2004). 8. Perry Bacon, Jr., “For Democrats, Pragmatism on Universal Health Care,” Washington Post, July 10, 1. 9. Of course, since low-income children live in multimember households, those choices

330

/

The Politics of Low-Income Families

10.

11.

12.

13. 14.

15.

16. 17. 18.

19.

20.

21.

22.

may not always be made with the objective of maximizing the child(ren)’s welfare. For a discussion, see Janet Currie (1997, 117–18). Southern states were also racially discriminatory in program administration during the early years of the program. On racial discrimination in the first decades of the AFDC and Old Age Assistance programs, see, for example, Jill Quadagno (1988), Robert C. Lieberman and John S. Lapinski (2001), and Lieberman (1998). See R. Shep Melnick (1994, 119). Indeed, membership on the Ways and Means subcommittee tended to be dominated by the committee’s most junior members, who switched off as soon as a more desirable subcommittee assignment became available. See Randall Strahan and R. Kent Weaver (1989). Two exceptions were the extension of Aid to Dependent Children benefits to a caretaker parent in 1950 and creation in 1962 of a state option to cover two-parent families where both parents were unemployed. The phrase is Gilbert Steiner’s (1971). Outlay figures are from Office of Management and Budget, Budget of the United States Government, Fiscal Year 2009, Historical Tables, p. 230. Tax expenditures figures are from Office of Management and Budget, Budget of the United States Government, Fiscal Year 2009, Analytical Perspectives, p. 290. Available at: http://www.gpoaccess.gov/usbud get/fy09/pdf/spec.pdf (accessed September 30, 2008). The expenditure data are from the Social Security Bulletin, Annual Statistical Supplement 2006, table 8.E2, and Office of Management and Budget, Budget of the United States Government, Fiscal Year 2008, Historical Tables, table 11-2. Available at: http://www .gpoaccess.gov/usbudget/fy09/pdf/hist.pdf (accessed September 30, 2008). See also Rebecca Adams, “Kids’ Health Gets Political,” CQ Weekly, July 23, 2007, 2179–84. Christopher Lee, “Bush’s Second Child Insurance Veto Stands in House,” Washington Post, January 24, 2008, A4. In addition, the Food, Conservation, and Energy Act of 2008, enacted over President Bush’s veto, removed limits on child-care deductions from income in determining Food Stamps eligibility. Deborah Stone (1989) stresses the importance of both causal ideas and the choices made by political actors in developing causal stories “in ways calculated to gain support for their side” (p. 282). A broader set of arguments about the role of values in determining American social policies focuses on attitudes toward state intervention generally and toward redistribution in general. But as Heclo (1986) has pointed out, American attitudes toward social policies have proved to be quite flexible, and some social programs, notably Social Security and Medicare, have been exceptionally popular. See also Quadagno and Debra Street (2005). On racialized images of poverty and welfare, see in particular Gilens (1999), especially chapters 5 and 6. See also Heclo (2001); Rosalee A. Clawson and Rakuya Trice (2000); and Sanford F. Schram, Soss, and Richard Fording (2003). See Philip Joyce (2005) and Roy T. Meyers and Joyce (2005). As Meyers and Joyce note, Congress failed to enact a budget resolution for the 1999, 2003, and 2005 fiscal years.

/

331

Making the Work-Based Safety Net Work Better 23. See, for example, Jonathan Weisman, “Reagan Policies Gave Green Light to Red Ink,” Washington Post, June 9, 2004, A11. 24. Paul D. Pierson (2001a, 2001b). Projections in the federal budgets prepared by the Bush administration that fiscal pressures will ease substantially in the next half decade are predicated on what appear to be unrealistically optimistic expectations about defense expenditures, the general funds deficit, and the expiration of tax cuts. 25. Whether this polarization is mirrored at the level of the electorate and whether high levels of congressional polarization are the norm or the exception are much more disputed. On electoral polarization see Morris P. Fiorina, Samuel J. Abrams, and Jeremy C. Pope (2006). On the polarization of congressional parties over time, see also Hahn and Brady (2007). 26. In U.S. Department of Labor, Employment and Training Administration (1998). 27. For example, Michael A. Bailey (2005, 134) argues that “welfare benefits exert a nontrivial effect on state residential choice” but those effects are dwarfed by the effect of poor single mothers returning to the state of their birth. See also Schram, Nitz, and Krueger (1998). 28. One of the few studies of AFDC to look at policy choices (for example, the option to cover children up through age twenty-one) is Russell L. Hanson (1983). 29. On the role of business in American social policy, and in particular the role played by small business, see Cathie J. Martin (1999). On business in the formative period the U.S. welfare state, see also Hacker and Pierson (2002). 30. For a discussion, see, for example, Public Sector Consultants, Inc. (2002). 31. Indeed, Mark Peffley and Jon Hurwitz (2007) found that whites who were exposed to the argument that the death penalty was primarily applied to blacks were actually more likely to favor the death penalty than those who were not exposed to that argument. 32. Francis Fox Piven and Richard A. Cloward (1971). See also the discussion in Walter I. Trattner (1983) and Fred Block (2003). 33. John Gilmour (1995). For an application to welfare reform politics, see Weaver (2000). 34. Joe Soss, personal communication, at the Institute for Research on Poverty Working conference, “Pathways to Self-Sufficiency: Getting Ahead in an Era Beyond Welfare Reform,” University of Wisconsin, Madison (September 7, 2007).

REFERENCES Adams, Rebecca. 2007. “Kids’ Health Gets Political.” CQ Weekly, July 23, 2179–84. Bailey, Michael A. 2005. “Welfare and the Multifacted Decision to Move.” American Political Science Review 99(1): 125–35. Berry, William D., Richard C. Fording, and Russell L. Hanson. 2003. “Reassessing the ‘Race to the Bottom’ in State Welfare Policy.” Journal of Politics 65(2): 327–49. Block, Fred. 2003. “Organizing versus Mobilizing: Poor People’s Movements After 25 years.” Perspectives on Politics 1(4): 733–35. Brown, Lawrence D., and Michael S. Sparer. 2001. “Window Shopping: State Health Reform Politics in the 1990s.” Health Affairs 20(1): 50–67. Brownlee, W. Elliot. 1996. Funding the Modern American State, 1941–1995: The Rise and Fall of 332

/

The Politics of Low-Income Families the Era of Easy Finance. Washington, D.C., and Cambridge: Woodrow Wilson Center for the Press and Cambridge University Press. Bucci, Michelle C., and William W. Beach. 2007. “22 Million New Smokers Needed: Funding SCHIP Expansion with a Tobacco Tax.” Heritage Foundation Web Memo No. 1548, July 11. Available at: http://www.heritage.org/research/healthcare/wm1548.cfm (accessed March 25, 2009). Burke, Vee. 2003. “Cash and Noncash Benefits for Persons with Limited Income: Eligibility Rules, Recipient and Expenditure Data, FY2000–FY2002.” CRS report, order code RL3223. Washington, D.C.: Congressional Research Service. Available at: http://www .opencrs.com/document/RL32233 (accessed March 25, 2009). Clawson, Rosalee A., and Rakuya Trice. 2000. “Poverty as We Know It: Media Portrayals of the Poor.” Public Opinion Quarterly 64(1): 53–64. Currie, Janet. 1997. “Choosing Among Alternative Programs for Poor Children.” The Future of Children 7(2): 113–31. Derthick, Martha. 1979. Policymaking for Social Security. Washington, D.C.: Brookings Institution. ———. 1990. Agency Under Stress: The Social Security Administration in American Government. Washington, D.C.: Brookings Institution. DiJulio, Bianca, and Paul D. Jacobs. 2007. Change in Percentage of Families Offered Coverage at Work, 1998–2005. Menlo Park, Calif.: Kaiser Family Foundation, July. Elving, Ronald D. 1995. Conflict and Compromise: How Congress Makes the Law. New York: Simon & Schuster. Edwards, George. 2008. Governing by Campaigning. New York: Pearson. Fellowes, Matthew C., and Gretchen Rowe. 2004. “Politics and the New American Welfare States.” American Journal of Political Science 48(2): 362–73. Ferejohn, John. 1986. “Logrolling in an Institutional Context: A Case Study of Food Stamps Legislation.” In Congress and Policy Change, edited by Gerald Wright, Leroy Rieselbach, and Lawrence Dodd. New York: Agathon Press. Fiorina, Morris P., Samuel J. Abrams, and Jeremy C. Pope. 2006. Culture War? The Myth of a Polarized America. 2nd ed. New York: Pearson Longman. Fording, Richard C., Joe Soss, and Sanford F. Schram. 2007. “Devolution, Discretion, and the Effect of Local Political Values on TANF Sanctioning.” Social Service Review 81(2): 285–316. Gais, Thomas, and R. Kent Weaver. 2002. “State Policy Choices Under Welfare Reform.” In Welfare Reform and Beyond, edited by Ron Haskins, Andrea Kane, Isabel Sawhill, and R. Kent Weaver. Washington, D.C.: Brookings Institution. Gilens, Martin. 1999. Why Americans Hate Welfare: Race, Media, and the Politics of Antipoverty Policy. Chicago: University of Chicago Press. Gilmour, John. 1995. Strategic Disagreement: Stalemate in American Politics. Pittsburgh: University of Pittsburgh Press. Gordon, Colin. 2004. Dead on Arrival: The Politics of Health Care in Twentieth Century America. Princeton, N.J.: Princeton University Press. Hacker, Jacob S. 2002. The Divided Welfare State: The Battle over Public and Private Social Benefits in the United States. Princeton, N.J.: Princeton University Press. ———. 2005. “Bringing the Welfare State Back In: The Promise (and Perils) of the New Social Welfare History.” Journal of Policy History 17(1): 125–54. /

333

Making the Work-Based Safety Net Work Better ———. 2006. The Great Risk Shift: The Assault on American Jobs, Families, Health Care, and Retirement—and How You Can Fight Back. New York and Oxford: Oxford University Press. Hacker, Jacob S., and Paul Pierson. 2002. “Business Power and Social Policy: Employers and the Formation of the American Welfare State.” Politics & Society 30(2): 277–325. ———. 2005. Off Center: The Republican Revolution and the Erosion of American Democracy. New Haven, Conn.: Yale University Press. Hahn, Hahrie, and David Brady. 2007. “A Delayed Return to Historical Norms: Congressional Party Polarization After the Second World War.” British Journal of Political Science 37(3): 505–31. Hanson, Russell L. 1983. “The Content of Welfare Policy: The States and Aid to Families with Dependent Children.” Journal of Politics 45(3): 771–85. Harrison, Kathryn. 1996. Passing the Buck: Federalism and Canadian Environmental Policy. Vancouver: University of British Columbia Press. Haskins, Ron. 2006. Work over Welfare: The Inside Story of the 1996 Welfare Reform Law. Washington, D.C.: Brookings Institution. Heclo, Hugh. 1986. “The Political Foundations of Antipoverty Policy.” In Fighting Poverty: What Works and What Doesn’t, edited by Sheldon H. Danziger and Daniel H. Weinberg. Cambridge, Mass.: Harvard University Press. Heclo, Hugh. 1994. “Poverty Politics.” In Confronting Poverty: Prescriptions for Change, edited by Sheldon H. Danziger, Gary D. Sandefur, and Daniel H. Weinberg. New York: Russell Sage Foundation and Harvard University Press. Heclo, Hugh. 2001. “The Politics of Welfare Reform.” In The New World of Welfare, edited by Rebecca M. Blank and Ron Haskins. Washington, D.C.: Brookings Institution. Howard, Christopher. 1997. The Hidden Welfare State: Tax Expenditures and Social Policy in the United States. Princeton, N.J.: Princeton University Press. ———. 2008. The Welfare State Nobody Knows: Debunking Myths About US Social Policy. Princeton, N.J.: Princeton University Press. Imig, Doug. 2006. “Building a Social Movement for America’s Children.” Journal of Children and Poverty 12(1): 21–37. Jacobs, Lawrence R., and Robert Y. Shapiro. 2000. Politicians Don’t Pander: Political Manipulation and the Loss of Democratic Responsiveness. Chicago: University of Chicago Press. Jenson, Jane. 2004. “Changing the Paradigm: Family Responsibility or Investing in Children.” Canadian Journal of Sociology 29(2): 169–92. Jenson, Jane, and Denis Saint-Martin. 2003. “New Routes to Social Cohesion? Citizenship and the Social Investment State.” Canadian Journal of Sociology 28(1): 77–99. Joyce, Philip G. 2005. “Federal Budgeting After September 11th: A Whole New Ballgame, or Is It Déjà Vu All Over Again?” Public Budgeting and Finance 25(1): 15–31. Kamerman, Sheila B., and Shirley Gatenio. 2002. “Tax Day: How Do America’s Child Benefits Compare?” Issue Brief. New York: Columbia University, Clearinghouse on International Developments in Child, Youth and Family Policies. Karch, Andrew. 2007 Democratic Laboratories: Policy Diffusion Among the American States. Ann Arbor: University of Michigan Press. Katz, Michael. 1987. In the Shadow of the Poorhouse. New York: Basic Books. Kaiser Commission on Medicaid and the Uninsured. 2007. Health Insurance Coverage in America: 2005 Data Update. Menlo Park, Calif.: Kaiser Family Foundation, May.

334

/

The Politics of Low-Income Families Kazee, Nicole. 2007. “The Invisible Variable: Bringing Business Back into State Antipoverty Policymaking.” Paper prepared for the annual conference on State Policy and Policymaking, Austin, February 23–24. Keiser, Lael R., Peter R. Mueser, and Seung-Whan Choi. 2004. “Race, Bureaucratic Discretion, and the Implementation of Welfare Reform.” American Journal of Political Science 48(2): 314–27. Kingdon, John W. 1995. Agendas, Alternatives, and Public Policies. 2d ed. New York: HarperCollins. Kletzer, Lori G., and Howard F. Rosen. 2006. Reforming Unemployment Insurance for the Twenty-First Century Workforce. Hamilton Project Discussion Paper 2006–06. Washington, D.C.: Brookings Institution, September. Lavery, Joni, and Virginia P. Reno. 2008 “Children’s Stake in Social Security.” Social Security Brief. Washington, D.C.: National Academy of Social Insurance. February. Leman, Christopher. 1977. “Patterns of Policy Development: Social Security in the United States and Canada.” Public Policy 25(Spring): 261–91. Lewin Group. 2008. McCain and Obama Health Care Policies: Cost and Coverage Compared. Falls Church, Va.: October 8. Lieberman, Robert C. 1998. Shifting the Color Line: Race and the American Welfare State. Cambridge, Mass.: Harvard University Press. Lieberman, Robert C., and John S. Lapinski. 2001. “American Federalism, Race and the Administration of Welfare.” British Journal of Political Science 31(2): 303–29. Lynn, Lawrence, and David Whitman. 1981. The President as Policymaker: Jimmy Carter and Welfare Reform. Philadelphia: Temple University Press. Mann, Cindy, Diane Rowland, and Rachel Garfield. 2003. “Historical Overview of Children’s Health Care Coverage.” The Future of Children 13(1): 31–52. Marmor, Theodore R. 1973. The Politics of Medicare. New York: Aldine, 1973. Martin, Cathie J. 1999. Stuck in Neutral: Business and the Politics of Human Capital Investment Policy. Princeton, N.J.: Princeton University Press. McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, Mass.: MIT Press. Melnick, R. Shep. 1994. Between the Lines: Interpreting Welfare Rights. Washington, D.C.: Brookings Institution. Mettler, Suzanne, and Joe Soss. 2004. “The Consequences of Public Policy for Democratic Citizenship: Bridging Policy Studies and Mass Politics.” Perspectives on Politics 2(1): 55–73. Meyers, Roy T., and Philip G. Joyce. 2005. “Congressional Budgeting at Age 30: Is It Worth Saving?” Public Budgeting & Finance 25(December): 68–82. Moynihan, Daniel P. 1973. The Politics of a Guaranteed Income; The Nixon Administration and the Family Assistance Plan. New York: Random House. Murray, Charles. 1984. Losing Ground: American Social Policy, 1955–1980. New York: Basic Books. National Conference of State Legislatures. n.d. “Universal Coverage Bills, 2006.” Available at: http://www.ncsl.org/programs/health/universalhealth06.htm (accessed August 21, 2007). Oberlander, Jonathan. 2003. The Political Life of Medicare. Chicago: University of Chicago Press.

/

335

Making the Work-Based Safety Net Work Better Oliver, Thomas R., Philip R. Lee, and Helene L. Lipton. 2004. “A Political History of Medicare and Prescription Drug Coverage.” Milbank Quarterly 82(2): 283–354. Pal, Leslie A. 1985. “Maternity Benefits and Unemployment Insurance: A Question of Policy Design.” Canadian Public Policy 11(3): 551–60. Patterson, James T. 1986. America’s Struggle Against Poverty, 1900–1985. Cambridge, Mass.: Harvard University Press. Peffley, Mark, and Jon Hurwitz. 2007. “Persuasion and Resistance: Race and the Death Penalty in America.” American Journal of Political Science 51(4): 996–1012. Peterson, Mark A. 1998. “The Politics of Health Care Policy: Overreaching in an Age of Polarization.” In The Social Divide, edited by Margaret Weir. Washington, D.C.: Brookings Institution, 1998 Peterson, Paul E., and Mark C. Rom. 1990. Welfare Magnets: A New Case for a National Standard. Washington, D.C.: Brookings Institution. Peterson. Paul E., Mark Carl Rom, and Kenneth F. Scheve, Jr. 1998. “Interstate Competition and Welfare Policy.” Publius 28(3): 17–37. Pierson, Paul D. 1994. Dismantling the Welfare State? Reagan, Thatcher and the Politics of Retrenchment. Cambridge and New York: Cambridge University Press. ———. 1995. “The Creeping Nationalization of Income Transfers in the United States, 1935–94.” In European Social Policy, edited by Stephan Leibfried and Paul Pierson. Washington, D.C.: Brookings Institution. ———. 2001a. “Coping with Permanent Austerity: Welfare State Restructuring in Affluent Democracies.” In The New Politics of the Welfare State. New York and Oxford: Oxford University Press. ———. 2001b. “From Expansion to Austerity: The New Politics of Taxing and Spending.” In Seeking the Center: Politics and Policymaking at the New Century, edited by Martin A. Levin, Marc K. Landy, and Martin Shapiro. Washington, D.C.: Georgetown University Press. Piven, Francis Fox, and Richard A. Cloward. 1971. Regulating the Poor: The Functions of Public Welfare. New York: Vintage Books. Public Agenda. 1999. “Kids These Days ’99.” Available at: http://www.publicagenda.org/ files/pdf/kids_these_days_99.pdf. Public Sector Consultants, Inc. 2002. “Framing Messages to Support Low-Income Families.” Report prepared for the Devolution Initiative of the W. K. Kellogg Foundation. Battle Creek, Mich.: December 4. Quadagno, Jill. 1988. “From Old-Age Assistance to Supplemental Security Income: The Political Economy of Relief in the South, 1935–1972.” In The Politics of Social Policy in the United States, edited by Margaret Weir, Ann Shola Orloff, and Theda Skocpol. Princeton, N.J.: Princeton University Press. Quadagno, Jill. 2005. One Nation, Uninsured: Why the United States Has No National Health Insurance. Oxford: Oxford University Press. Quadagno, Jill, and Debra Street. 2005. “Ideology and Public Policy: Antistatism in American Welfare State Transformation.” Journal of Policy History 17(1): 52–71. Riccucci, Norma M. 2005. “Street-Level Bureaucrats and Intrastate Variation in the Implementation of Temporary Assistance for Needy Families Policies.” Journal of Public Administration Research and Theory 25(1): 89–111.

336

/

The Politics of Low-Income Families Sawhill, Isabel V., ed. 2003. One Percent for the Kids: New Policies, Brighter Futures for America’s Children. Washington, D.C.: Brookings Institution. Schram, Sanford, Lawrence Nitz, and Gary Krueger. 1998. “Without Cause or Effect: Reconsidering Welfare Migration as a Policy Problem.” American Journal of Political Science 42(1): 210–30. Schram, Sanford F., and Joe Soss. 1998. “Making Something out of Nothing: Welfare Reform and a New Race to the Bottom.” Publius 28(3): 67–78. Sanford F. Schram, Joe Soss, and Richard Fording, eds. 2003. Race and the Politics of Welfare Reform. Ann Arbor: University of Michigan Press. Schram, Sanford F., Joe Soss, Thomas P. Vartanian, and Erin O’Brien. 2001. “Setting the Terms of Relief: Explaining State Policy Choices in the Devolution Revolution.” American Journal of Political Science 45(2): 378–95. Sherman, Arloc. 2005. “Social Security Lifts 1 Million Children Above the Poverty Line.” May 2. CBPP Policy brief. Washington D.C.: Center on Budget and Policy Priorities. Available at: http://www.cbpp.org/archiveSite/5-2-05socsec.pdf (accessed March 25, 2009). Skocpol, Theda. 1991. “Targeting Within Universalism: Politically Viable Policies to Combat Poverty in the United States.” In The Urban Underclass, edited by Christopher Jencks and Paul E. Peterson. Washington, D.C.: Brookings Institution. Soss, Joe. 1999. “Lessons of Welfare: Policy Design, Political Learning, and Political Action.” American Political Science Review 93(2): 363–80. Soss, Joe, and Sanford Schram. 2007. “A Public Transformed? Welfare Reform as Policy Feedback.” American Political Science Review 101(1): 111–27. Sparer, Michael S. 2004. “States and the Politics of Incrementalism: Health Policy in Wisconsin During the 1990s.” Journal of Health Politics Policy and Law 29(2): 269–92. Steiner, Gilbert Y. 1971. The State of Welfare, Washington, D.C: Brookings Institution. Steiner, Gilbert Y. 1976. The Children’s Cause. Washington, D.C.: Brookings Institution. Steinmo, Sven, and Jon Watts. 1995. “It’s the Institutions, Stupid! Why Comprehensive National Health Insurance Always Fails in America.” Journal of Health Politics, Policy and Law 20(Summer): 329–72. Steuerle, C. Eugene. 2003. The Incredible Shrinking Budget for Working Families and Children. National Budget Issues Brief No. 1. Washington, D.C.: Urban Institute, December. Stone, Deborah A. 1989. “Causal Stories and the Formation of Policy Agendas.” Political Science Quarterly 104(2): 281–300. Strahan, Randall, and R. Kent Weaver. 1989. “Subcommittee Government and the House Ways and Means Committee.” Paper presented at the annual meeting of the Southern Political Science Association. Memphis (November 2 to 4). Trattner, Walter I., ed. 1983. Social Welfare or Social Control? Some Historical Reflections on “Regulating the Poor.” Knoxville: University of Tennessee Press. Tweedie, Jack. 1994. “Resources Rather Than Needs: A State-Centered Model of Welfare Policy-Making.” American Journal of Political Science 38(3): 651–72. U.S. Department of Labor, Employment and Training Administration. 1998. Essays on Interstate Competition in the Unemployment Insurance System. Unemployment Insurance Occasional Paper 98-5. Sections I and III. Washington, D.C.: U.S. Department of Labor, Employment and Training Administration. Available at: http://wdr.doleta.gov/owsdrr/ 98-5/98-5.pdf.

/

337

Making the Work-Based Safety Net Work Better U.S. Office of Management and Budget. 2007. Budget of the United States Government, Fiscal Year 2008, Analytical Perspectives. Washington, D.C.: U.S. Office of Management and Budget. Volden, Craig. 2002. “The Politics of Competitive Federalism: A Race to the Bottom in Welfare Benefits?” American Journal of Political Science 46(2): 352–63. Weaver, R. Kent 1996. “Deficits and Devolution in the 104th Congress.” Publius 26(3): 45–85. ———. 2000. Ending Welfare As We Know It. Washington, D.C.: Brookings Institution. ———. 2005a. “Public Pension Policy in the United States.” In Ageing and Pension Reform Around the World: Evidence from Eleven Countries, edited by Giuliano Bonoli and Toshimatsu Shinkawa. Cheltenham, U.K.: Edward Elgar. ———. 2005b. “Public Pension Reform in an Age of Austerity.” In Trans-Atlantic Policymaking, edited by Martin Levin and Martin Shapiro. Washington, D.C.: Georgetown University Press. Wilson, William Julius. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago: University of Chicago Press. Wright, Gerald C. 1977. “Racism and Welfare Policy in America.” Social Science Quarterly 57(4): 718–30.

338

/

Index Boldface numbers refer to figures and tables

Abecedarian, 132 Abt Associates, Inc., 21n17 academic achievement. See school achievement academies, career, 49, 57–59 accidents, and unemployment rate, 84 accountability, school, 163–67, 169–70 Acemoglu, D., 42 Acs, G., 20n9, 232, 270–71 Adams, S., 68n24 Adema, W., 264, 273 adolescents. See youth adult education programs, 12–13 AFDC (Aid to Families with Dependent Children), 69n34, 121, 140, 303–4, 319–20 AFQT (Armed Forced Qualifying Test), 201 African Americans: class size reduction impact, 155; discrimination, 331n10; incarceration rates, 185–86, 189–95; school-to-work program participation, 71n45; school voucher program impact, 170 age analyses, 17, 119–20, 209, 210 Aid to Families with Dependent Children (AFDC), 69n34, 121, 140, 303–4, 319–20 air pollution, 84 alcohol consumption, 82, 84 Allegretto, S., 67n11 America’s Second Harvest, 255n8 Anderson, P., 166 anemia, 88–89, 97, 98 anti-poverty program effects, 29–36, 40–41, 44–46, 61 Apel, R., 203 apprenticeships, 49, 53, 57 Armed Forced Qualifying Test (AFQT), 201 Asian Americans, incarceration rates, 189, 190

assets, 13–14 Assets for Independence, 13 asset tests, 302 background checks, 199, 217 backlash politics, 311 Bailey, M., 332n27 Baker, D., 84 Barrow, L., 149 Bauman, K., 232 Beck, A., 221n4 behavior, of children, 124, 136, 186 Behavioral Risk Factor Surveillance Study (BRFSS), 83 benefits, welfare, 67n6, 245, 271, 302–4 Bergmann, B., 287–88n14 Berlin, G., 62, 63 Bernstein, J., 67n12 Bifulco, R., 173, 174 bipartisanship, 318 Bitler, M., 67n4 Black, W., 84 blacks. See African Americans Blank, R., 72n58, 232, 251–52 Bloom, D., 198, 218, 222n13 Blumstein, A., 221n4 body fat tests, 87, 95 body mass index (BMI), 87, 94, 95 bonuses, 158 Boozer, M., 154 Bos, J., 136 Boyd, D., 161, 166 Brame, R., 218, 219 Brenner, M., 68n24, 110n3 BRFSS (Behavioral Risk Factor Surveillance Study), 83 Brown, C., 28 Brown, L., 320 Brownlee, E., 312 budget deficits, national, 313, 314 Bureau of Justice Statistics, 185–86, 220n1 Bureau of Labor Statistics, 110n6, 262 Burkhauser, R., 37, 250

Bush (George H. W.) administration, 163 Bush (George W.) administration, 306, 312, 328 Bushway, S., 218, 219 business community, 322 Butcher, K., 166, 192 California: class size reductions, 155; earnings of welfare leavers, 4; Greater Avenues to Independence, 1, 2–3, 125, 141; incarceration and employment consequences, 197–98, 204; incarceration rates, 192, 221n2; living wage laws, 39; Proposition 13, 312 California Department of Corrections, 192 Canada: maternity benefits, 301; Self-Sufficiency Program, 125, 135, 141 cancer, 81 car accidents, and unemployment rate, 84 Card, D., 28, 37, 42, 154 cardiovascular disease, 81, 84, 88 career academies, 49, 57–59 Carruthers, C., 173 Carter administration, 303 caseloads, 2, 249 cash assistance, 302–4. See also Temporary Assistance for Needy Families (TANF) Cauthen, N., 287n5 census data, 149, 189, 192, 205 center-based child care, 132–34 Center for Employment Opportunities (CEO), 198, 218 Center on Budget and Policy Priorities, 251, 300, 327 Center on Law and Social Policy, 327 Centers for Disease Control (CDC), National Health and Nutrition Examination Survey (NHANES), 80, 86–87

Index CEO (Center for Employment Opportunities), 198, 218 Chakrabarti, R., 171 Chaplin, D., 42 Charlotte-Mecklenburg public schools, 164 charter schools, 168–69, 173–75 Chay, K., 84 Chicago, Ill., living-wage law, 39 Chicago Public Schools, 163–64, 165, 166 child and family allowances, 271, 273, 280 child care: relative-provided, 247, 278; and school achievement, 132–34, 136–37; subsidized programs, 264, 267, 276–79, 280, 296, 298 child development, 119, 123 child poverty, 230, 264, 279, 300 children, as social investment, 325. See also youth children’s well-being: data sources, 118, 120–22; differences across age span, 119–20; and income, 119; and maternal employment, 119; methodology, 117–18, 121–22; poverty rates, 230, 264, 279, 300; research considerations, 118. See also school achievement child support, 248, 273, 325 Child Tax Credit (CTC), 271, 306, 314, 318, 328 Chile, school vouchers in, 172 Chin, M., 247 Cho, R., 204 cholesterol levels, 88, 94, 96 Chou, S., 83 chronic illnesses, 79, 82, 87, 88 churning, 47, 49 Citro, C., 5 Clark-Kauffman, E., 123 class size, 154–56, 176 Cleveland, Ohio, school voucher program, 171 Clinton administration: education policy, 163; EITC expansion, 305; health care policy, 302, 312; issue-framing, 324–25; Medicaid, 307; welfare reform, 304 Clotfelter, C., 158, 159 Cloward, R., 332n32 coalitions, 302 cognitive development, 119 cohabitation, 231–32 Cohen, W., 327

340

/

college education and graduates: children’s school readiness and achievement, 150, 151; earnings, 149; incarceration rates, 190; and school-towork programs, 52–53, 58–59; unemployment and health status, 85 Columbia University Social Indicators Survey, 266 community-based organizations, 246–47 community colleges, school-towork programs, 59–60 compensation, teachers, 157–60 compromise, 327 Congress, 305, 307–8, 314, 315–18, 327. See also specific laws Congressional Research Service, 294 Connecticut Jobs-First Evaluation, 125, 141 conservatism, 294, 306 consumption, of poor families, 231 contractors, city, 36, 39 Cook, D., 81, 82 coop programs, 53 corporate responsibility, 66n2 corporations, hostility to safety net, 322 cost-benefit analyses, 135 counterfactuals, 201 courts, 294 CPS (Current Population Survey), 30, 32, 222n12, 232–36, 247 credentialing, of teachers, 161 crime, 18–19, 63, 206, 211–15. See also incarceration criminal background checks, 199, 217 criminal history records, 186, 198–99, 217, 218–20 Crosby, D., 132 cross-national comparison, of social benefits package: child care, 276–79, 285–86; definition of, 263–64; distribution and structure of, 266–68; family leave, 274–76; government funding, 264–65, 283, 284–85; income benefits, 269–74; policy recommendations, 280–83; research considerations, 261–62; U.K. programs, 280–81 CTC (Child Tax Credit), 271, 306, 314, 318, 328

Current Population Survey (CPS), 30, 32, 222n12, 232–36, 247 Currie, J., 231, 267 Dahl, G., 152 Dallas public schools, 164 Daly, M., 250 Danielson, C., 255n5 Danziger, S., 232, 255n3 Darling-Hammond, L., 162 data sources: children’s wellbeing, 118, 120–22; disconnected single mothers, 232–33; health, 80, 86–87; work support programs, 120–22. See also specific sources DCTC (Dependent and Child Care Tax Credit), 277 death penalty, 332n31 Dee, T., 159 Deficit Reduction Act (2005), 2 deficits, national, 313, 314, 315 DeLeire, T., 14 Democratic Party, 307–8, 314, 315–18 Denmark: child and family allowances, 273; child care, 278; child poverty, 264; family leave, 275, 276; Unemployment Insurance, 272, 273 Department of Labor, 110n6 Dependent and Child Care Tax Credit (DCTC), 277 depression, 236 deterrence effects, of incarceration, 211–14, 216 Detroit, Mich., living-wage law, 39 development, child, 119, 123 diabetes, 88, 94, 97 Dickert-Conlin, S., 46, 47, 48, 62, 70n35 difference-in-differences approach, 32–33, 42 disabilities, people with, SSI receipt, 237, 243, 245, 250–51, 303 disadvantaged youth, schoolto-work program impact, 53, 55–57 disconnected single mothers: characteristics of, 236–40; data sources, 232–33; definition and measurement of, 232–36; financial and nonfinancial resources, 244–48; policy considerations, 244, 248–53; research considerations, 228, 254; statistics,

Index 227–28; time period of disconnectedness, 240–44 discrimination, 199–200, 216–17, 219, 331n10 disemployment effects, 28–29 distributional effects, 29–36, 40–41, 44–46, 61 domestic violence, 236 Donohue, J., 215 Dorrance, J., 62 drug offenders, 208, 214, 216 Dube, A., 67n11, 287n14 Duncan, G., 123 early childhood education: crime prevention impact, 215; expansion in U.S., 277; for low-income families, 277–78; in Nordic countries, 278–79; research support, 17; and school readiness, 150; universal program cost, 285–86 Early Childhood Longitudinal Program-Kindergarten (ECLS-K), 150–51 Earned Income Tax Credit (EITC): current policy discussions, 62–64; distributional effects, 44–46; vs. employer-based wage subsidies, 47, 48; employment effects, 43–44; expansion of, 2, 43, 305–6, 315; extension to unrelated individuals, 45, 62–64; origins of, 305; payment determination, 43, 271; politics of, 305–6; state supplements, 69n28, 271; supply-and-demand analysis, 61; and working poor, 267 earnings and wages: after incarceration, 201–5; and educational attainment, 149; minimum wage effects, 30, 32, 43; policy interventions, 25–26; of single mothers, 230–31; student-teacher ratio impact, 154; of welfare leavers, 3–4; welfare reform impact, 3–4 earnings-supplement programs, 121, 124, 129–30, 135–36, 139–40 eating habits, 83 EBT (Electronic Benefits Transfer), 308 ECLS-K (Early Childhood Longitudinal Program-Kindergarten), 150–51

economic security, 263, 324 economic stimulus package, 312, 314, 328 economy, 82–84, 292, 310, 311 Edin, K., 246, 248 Educate America Act, 163 educational attainment: and children’s school readiness and achievement, 130, 132, 139, 149–51; and earnings, 26, 149; and incarceration rates, 189–91; minimum wage effects, 42; rate of return for extra year of schooling, 150; and socioeconomic status, 150; of welfare recipients, 2 Education Department, U.S., 159 Education Longitudinal Study (2000), 151 education reform. See school reform EITC (Earned Income Tax Credit). See Earned Income Tax Credit (EITC) elderly, 85 Electronic Benefits Transfer (EBT), 308 elementary school children, academic achievement, 123 eligibility tests, 267 elite, dissensus of, 323 employer-based benefits: health insurance, 245, 246, 265, 268, 271, 309; in U.S. vs. OECD countries, 283 employer-based wage subsidies, 46–48, 62 employment: barriers for welfare leavers, 236–40, 249, 251–52; and children’s school achievement, 126–30; and EITC, 43–44; of high school dropouts, 1, 2; incarceration effects, 195–205; and income supplements, 270, 284; living wage effects, 39; minimum wage effects, 27–29, 61, 101; of mothers, 119; part-time status, 272–73; policy considerations, 15–16; school-towork program impact, 51–60. See also unemployment employment bans, against exoffenders, 216–17 Emsellem, M., 287n13 endogenous selection, 51–52, 55 energy assistance, 296, 298

Engelhardt, G., 13–14 Epple, D., 172 equality of opportunity, 310 Erikson, E., 107 ES-202 earnings records, 204 Ettner, S., 84–85 Evans, W., 85 ex-offenders, disqualification from public assistance, 186, 216. See also incarceration expenditures, government. See government funding expenditures per pupil, 153 experiments, state-level, 318–19 Fairris, D., 41, 42, 68n24 family, financial assistance from, 247 Family and Medical Leave Act (FMLA) (1993), 274–75 Family Assistance Plan, 303, 305 family caps, 304, 320 family income. See income family leave, 263, 274–76, 280, 285–86 family structure, 14–15 Family Support Act (1988), 121 Family Transition Program (FTP), 125, 141 fast-food industry, 28, 83 fathers, child support, 248, 273, 325 federalism, 19, 318–22 Feeding America, 255n8 Figlio, D., 72n56, 154, 156, 159, 166, 167, 169–70, 171, 172 financial aid, 216 Finland: child and family allowances, 273; child care, 278; family leave, 275, 276; Unemployment Insurance, 272 fiscal constraint, 311–15 Fiske, E., 172 flexible spending accounts, 277 flexible work schedules, 286n1 Florida: charter schools, 173, 174; employment consequences of incarceration, 204; school vouchers, 168, 169, 171–72; welfare time limits, 140 FMLA (Family and Medical Leave Act) (1993), 274–75 folate, 87–88, 94, 97, 98 food pantries, 246–47 food prices, 83

/

341

Index Food Stamps: drug felons’ ineligibility, 216; eligibility, 244–45; expansion of takeup, 250, 253; government spending on, 296, 298; lowincome single mothers’ receipt of, 246; politics of, 308; redistributional effects, 69n34; working poor families’ participation in, 267 foster care, 296, 298 Fowler, R. Clarke, 158 fragmentation, institutional, 315–18 framing, of issues, 310–11, 324–26 France, child poverty in, 264 Frank, R., 255n3 FTP (Family Transition Program), 125, 141 funding. See government funding GAIN (Greater Avenues to Independence), 1, 2–3, 125, 141 Gallo, W., 81 GAO (Government Accountability Office), 70n38 Gardiner, J., 5 Gatenio, S., 265 Gelbach, J., 67n4 gender equality, 275, 276 Gennetian, L., 123, 124, 132, 134 Gerdtham, U., 81, 83 Germany: apprenticeships, 49; unemployment and mortality rates, 83 Gilens, M., 310 Gilmour, J., 327 Gilroy, C., 28 Gitterman, D., 62 Glennie, E., 171 Glewwe, P., 159 Golan, A., 38, 45, 67n19, 69n34 Gorham, L., 62 Government Accountability Office (GAO), 70n38 government funding: policy recommendations, 283, 284–86; political issues, 302, 311–15; of specific programs as percentage of GDP, 295, 296–98; in U.S. vs. other OECD countries, 264–65 Gramlich, E., 32 Great Britain. See United Kingdom Greater Avenues to Independence (GAIN), 1, 2–3, 125, 141

342

/

Greene, J., 171 Greenstone, M., 84 Grogger, J., 3, 69n32, 204, 222n14 Grossman, M., 83 Grubb, W. Norton, 59 Gueron, J., 4 Gunderson, C., 38, 68n19, 69n34 Guryan, J., 153 Hacker, J., 309, 318 Haider, S., 247 Hamersma, S., 48, 70n38 Hamilton, V., 83 Hanratty, M., 255n5 Hanushek, E., 156, 157, 173 Hao, L., 247 Head Start, 123, 132, 277 health: cardiovascular disease, 81, 84, 88; causal effect of working on, 81, 98, 100–103, 105, 106–7; of children, 124; chronic illnesses, 79, 82, 87, 88; data sources, 80, 86–87; of disconnected single mothers, 236, 238; measurement of, 87–89; obesity, 83, 87, 94; policy considerations, 105–6; research considerations, 79–85, 107–8; trends, 94–98, 104–5; and unemployment, 79, 80, 81–85; of workers vs. nonworkers, 89–94, 98, 100 health care, 83, 302, 312 health insurance: Americans’ anxiety over, 309–10; employer-provided, 245, 246, 265, 268, 271, 309; lack of, 14, 98, 108, 309–10; research considerations, 14, 108; universal programs, 301, 320–21. See also Medicaid heart disease, 81, 84, 88 Heckman, J., 70n40 Heclo, H., 293, 295, 300, 331n20 Helburn, S., 287–88n14 “hidden welfare state,” 304–5 higher education. See college education and graduates high-risk youth, 205 high-road work practices, 66n2 High School and Beyond, 151 high school dropouts: children’s school achievement, 150–51; earnings, 149; employment rate, 1; incarceration rates, 190, 192; policy considerations, 205; poverty rates, 149

high school graduation rates, 59 high schools, school-to-work programs, 48–59 Hispanics, incarceration rates, 185–86, 189–95 HIV/AIDS, 186 Hoffman, S., 43 Hogan, D., 247 Holt, S., 268 Holtz-Eakin, D., 46, 47, 48, 62, 70n35 Holzer, H., 13, 17, 65, 199, 217, 221n9, 270–71 home-based child care, 132–34 home environment, 134–35 homicide, 84 Horrigan, M., 45 Hotz, V. Joseph, 43 household income. See income housing, 13 Housing and Urban Development (HUD), 13 Howard, C., 304–5 Howell, W., 170 Hoxby, C., 168, 171, 173 Hoynes, H., 67n4 human capital, 26–27, 152 human capital-development (HCD) approach, of NEWWS, 130, 132 Hurwitz, J., 332n31 Huston, A., 124 IDAs (Individual Development Accounts), 13 identity, 107 Ilias, N., 159 Illinois, employment consequences of incarceration, 204 Imig, D., 311 immigrants, incarceration rates, 192 incarceration: age analysis, 209, 210; average time spent in, 198; and crime rate, 206, 211–15; earnings impact, 201–4; employment consequences, 195–205; measurement of, 189; policies to limit adverse consequences of, 18–19, 186–87, 204–5, 215–20; reduction without crime rate increase, 205–10; social costs, 186, 214; trends, 185–95, 206, 214 incentives, for work, 26 income: and children’s school achievement, 120, 126–30, 134, 136–37, 152; and chil-

Index dren’s well-being, 119; EITC impact, 44; and health status, 84–85; minimum wage effects, 32–36; NHANES measure, 86; of single mothers, 230–31; welfare reform impact, 3 income supplements or benefits: business community’s support for, 322; definition of, 263; for disconnected single mothers, 251–52; in Nordic countries, 271–74; role of, 269; in U.K., 279–80; in U.S., 270–71; and work behavior, 270, 284 income-tested cash assistance, 302–4 income-tested in-kind assistance, 306–9 India, teacher merit pay in, 160 Individual Development Accounts (IDAs), 13 infants, 120, 137 inflation indexing, 314, 328 in-kind assistance, 306–9, 310. See also Food Stamps; Medicaid Institute for Research on Poverty (IRP), 3–4 institutional fragmentation, 315–18 instrumental-variables (IV), 130, 132, 134, 212 insurance, health. See health insurance intellectuals, 323 interdisciplinary approaches, 117 international study. See crossnational comparison, of social benefits package internships, 53, 57 IQ tests, 124 IRP (Institute for Research on Poverty), 3–4 Israel, teacher merit pay in, 159, 160 issue framing, 310–11, 324–26 Jacob, B., 166, 167 Jacobs, L., 325–26 Japan: school-to-work programs, 49; work hours and heart health, 84 Jenson, J., 325 Job Corps, 70n40 job loss. See unemployment Job Opportunities in the Business Sector (JOBS), 46

jobs, generally. See employment job search assistance, 217–18 job shadowing, 52 job training. See training programs Job Training and Partnership Act (JTPA), 46, 222n13 job turnover, 3 Johannesson, M., 81 Johnson, R., 212, 214 Johnson, W., 152 Johnson administration, 308 Joyce, P., 331n22 Jung, H., 204 junior high, transition to, 123 Kagamimori, S., 84 Kain, J., 156 Kamerman, S., 265 Kaplan, E., 287n14 Karasek, R., 84 Karoly, L., 3 Katz, L., 13, 46–47, 48 Kauff, J., 238 Kazee, N., 320 Kemple, J., 58, 59, 71n47 Kenney, C., 232 Kenny, L., 159 Kenya, teacher merit pay in, 159 Keys, B., 159 kindergarten readiness, 150, 277 King, C., 3 Kingdon, J., 330n2 Klerman, J., 255n5 Kling, J., 13, 204, 222n12 Kohen, A., 28 Kornfeld, R., 222n13 Korpi, T., 81 Kortenkamp, K., 255n3 Kremer, M., 159 Krueger, A., 28, 37, 154, 155, 170 Kurlycheck, M., 218, 219 Kuttner, H., 14 Labor Department, U.S., 110n6 labor-force attachment (LFA) approach, of NEWWS, 132 labor-force participation, 1, 15–16, 262 labor-market policies: current status and proposals, 60–66; employer-based wage subsidies, 46–48, 62; living wage laws, 36, 39–41, 61–62; research considerations, 25–27; school-to-work programs, 48–60, 61, 64–65. See also

Earned Income Tax Credit (EITC); minimum wage labor productivity, 26 labor supply, 26, 43–44 labor unions, 322 Ladd, H., 158, 159, 171, 172, 173, 174 Lakdawalla, D., 83 Lalonde, R., 70n40, 204 Laub, J., 222n14 Lavy, C., 159, 160 Lazear, E., 154, 158 learnfare program, 67n6 leave, family, 263, 274–76, 280, 285–86 Legal Action Center, 217 Lein, L., 246, 248 leisure time, 83, 105, 107–8 Lester, T. William, 67n11 Levine, J., 14 Levitt, S., 166, 222n15 Levy, H., 14 liberals, 315–18 Liebman, J., 13 LIHEAP (Low Income Home Energy Assistance Program), 296, 298 LIS (Luxembourg Income Study), 264 living expenses, 247 living wage laws, 36, 39–41, 61–62 lobbying, 301, 302, 322 local governments, living wage laws, 36, 39–41 Lochner, L., 152 Long, R., 305 Loprest, P., 232 Los Angeles, Calif.: Jobs-First Greater Avenues for Independence (GAIN), 1, 2–3, 125, 141; living-wage law, 39 Losing Ground (Murray), 304 Low Income Home Energy Assistance Program (LIHEAP), 296, 298 Lucas, M., 156 Luxembourg Income Study (LIS), 264 Lyons, C., 204 macroeconomics approach, to health and employment, 81, 82–84, 85 MaCurdy, T., 64 Madison, Wis., minimum wage, 68n20 Magnuson, K., 132, 247 maintenance-of-effort (MOE) requirements, 251, 253

/

343

Index Manpower Demonstration Research Corporation (MDRC), 20n11, 57–58, 198, 218 Manski, C., 108 marginal offenders, 206–14 market income, 261 marriage, 14–15, 62, 63, 186, 232 Marschall, M., 169 Massachusetts, school spending in, 153 maternity leave and benefits, 274–76, 280, 301 math, 151, 167 Mathematica Policy Research, 50 Mathers, C., 82 McCarty, N., 318 McGarry, K., 247 McLaughlin, C., 14 MDRC (Manpower Demonstration Research Corporation), 20n11, 57–58, 198, 218 Mead, L., 2, 20n6 means-tested public assistance programs, 250, 267–68, 273 means-tested voucher programs, 172 Meara, E., 255n3 Medicaid: disconnected single mothers, 245–46, 250; eligibility loss when employed, 98, 108; government spending on, 296, 298; politics of, 307; and working poor, 267 medical care, 83 Medicare, 267, 300–301, 302 Melnick, R. Shep, 331n11 men: health status, 91–93, 95–97, 99, 100, 102–3; incarceration rates, 185–86, 189–95; labor-force participation, 1; school-to-work participation effects, 56, 57, 58–59 mental illness, 236, 238 mentoring programs, 52 merit pay, 158–60, 175–76 methodology: causal effects of working on health, 101–3, 108–9; children’s well-being, 117–18, 121–22; employment consequences of incarceration, 200–205; work support programs, 120–22 Meyer, B., 231 Meyers, M., 266 Meyers, R., 331n22 MFIP (Minnesota Family Investment Program), 124, 125, 135, 141

344

/

Michael, R., 5 microeconomics approach, to health and employment, 81, 85 migration, welfare effect, 319 Miller, C., 124 Milwaukee, Wis.: New Hope Project, 63, 121, 124, 125, 136, 141; school voucher program, 171 Mincy, R., 45 minimum standard of living, 5 minimum wage: at city level, 68n20; distributional effects, 29–36, 45–46, 61; employment effects, 27–29, 61, 101; and skill acquisition, 41–43; supply-and-demand analysis, 61 Minnesota, barriers to employment for welfare leavers, 238 Minnesota Family Investment Program (MFIP), 124, 125, 135, 141 minorities, incarceration rates, 53, 185–86, 189–95. See also specific groups moderate legislators, 318 MOE (maintenance-of-effort) requirements, 251, 253 Morris, J., 81, 82, 123 Morris, P., 122 mortality, 81, 82–83 mortgage interest deduction, 304 mothers, 119, 230–32, 262, 300. See also disconnected single mothers Moving to Opportunity, 13 Mueser, P., 3 Multi-City Study of Urban Inequality, 199 Murarka, S., 173 Murray, C., 304 Myth and Measurement (Card and Krueger), 28 National Academy Foundation (NAF), career academies, 71n50 National Academy of Science, 137 National Assessment of Educational Progress (NAEP), 151 National Corrections Reporting Program (NCRP), 207–11 National Education Longitudinal Study (1988), 151 National Evaluation of Wel-

fare-to-Work Strategies (NEWWS), 124, 125, 132, 141 National Governor’s Association, 20n11 National Health and Nutrition Examination Survey (NHANES), 80, 86–87, 94, 104 national health insurance, 301, 320 National Longitudinal Survey of Youth (NLSY79), 201–3, 222n14 National Longitudinal Survey of Youth (NLSY97), 51–53 National Survey of America’s Families, Urban Institute, 14 NCLB (No Child Left Behind) Act, 49, 65, 163–64 NCRP (National Corrections Reporting Program), 207–11 Neal, D., 152, 165 negative income tax, 303 negligent hiring, 198 Neumark, D., 30, 32, 36, 42, 45, 46, 49, 53, 55, 68n24, 70n34, 43, 72n54, 58 Neumayer, E., 83 New Deal, 300 New Hope Project, 63, 121, 124, 125, 136, 141 New Jobs Tax Credit (NJTC), 46 New Orleans, La., minimum wage, 68n20 NEWWS (National Evaluation of Welfare-to-Work Strategies), 124, 125, 132, 141 New York: Center for Employment Opportunities, 198, 218; teacher credentials, 161; test-based education reform, 166 New York City: living-wage law, 39; Opportunity NYC program, 205–6; school vouchers, 170; social benefits package study, 266 New Zealand, universal public school choice, 172 Next Generation Project, 118, 120–22 NHANES (National Health and Nutrition Examination Survey), 80, 86–87, 94, 104 Nichols, A., 270–71 Nightingale, D. Smith, 13, 17 Nixon administration, 303, 305 Nizalova, O., 42, 72n54 NJTC (New Jobs Tax Credit), 46 NLSY79 (National Longitudi-

Index nal Survey of Youth), 201–3, 222n14 NLSY97 (National Longitudinal Survey of Youth), 51–53 No Child Left Behind (NCLB) Act, 49, 65, 163–64 Nordic model, 269, 271–76, 278–79 North Carolina: charter schools, 173, 174; school accountability, 171; teacher bonuses, 158 Norway: child and family allowances, 273; family leave, 275, 276 nutritional adequacy, 87–88 nutrition programs, 296, 298, 308 OASDI (Old Age, Survivors, and Disability Insurance), 281, 283, 299–300 Obama administration, 312, 314, 324, 328 Obama campaign and election, 292, 301 obesity, 83, 87, 94 OECD (Organization for Economic Cooperation and Development), 83, 261, 262, 264–66, 269 Ohio, employment consequences of incarceration, 204 Old Age, Survivors, and Disability Insurance (OASDI), 281, 283, 299–300 Opportunity NYC program, 205–6 ordinary least squares (OLS), 122, 134 Oreopoulos, P., 152 Organization for Economic Cooperation and Development (OECD), 83, 261, 262, 264–66, 269 Orr, M. Terry, 71n50 Orzol, S., 255n3 out-of-wedlock births, 62 Page, M., 152, 169–70 Pager, D., 199–200 paid family leave, 263, 274–76, 280, 285 Pape, A., 42 parenting, 119, 126, 134–35, 139, 311 parole violations, 206, 207–8, 214 Parrish, W., 247 partisanship, 315–18

part-time employment, 272–73 Pavetti, L., 20n9, 238 pay for performance programs, 158–60, 175–76 Pedace, R., 42 Peffley, M., 332n31 Pell grants, 216 pensions, 265, 271 Pepper, J., 108 Perloff, J., 38, 45, 67n19, 69n34 “permanent austerity,” 311–15 Perry Preschool, 132 Person, A., 59–60 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) (1996), 1–2, 121, 216, 310–11. See also Temporary Assistance for Needy Families (TANF) Peterson, P., 170 Pettit, B., 192, 204 phase-outs, benefit, 302 Philadelphia, Pa., living-wage law, 39 Philipson, T., 83 physical activity: causal effect of working on, 103; and economy, 83; measurement of, 87, 89; trends, 98, 99; of workers vs. nonworkers, 91–92 Piehl, A. Morrison, 192 Pierson, P., 311, 318, 332n24 Pischke, J., 42 Piven, F. Fox, 332n32 polarization, political, 315–18 policy: complexity of, 294; cross-national comparison lessons, 280–83; disconnected single mothers, 244, 248–53; federal vs. state responsibility, 19; government funding, 283, 284–86; health, 105–6; high school dropouts, 205; incarceration consequences, 18–19, 186–87, 204–5, 215–20; incremental vs. extensive changes, 17–18; integration of approaches, 15–16; political streams, 294–95; “propoor” agendas, 326–27; school reform, 162–63; strategic issues, 324–29; targeting to children vs. adults, 17; tradeoffs, 268–69; working poor, 268–69. See also labormarket policies political issues and constraints: business community’s hostility, 322; elite dissensus, 323;

federalism, 318–22; framing of issues, 310–11, 324–26; “great risk shift” factor, 309–10; income-tested cash assistance, 302–4; incometested in-kind assistance, 306–9; partisanship, 315–18; “permanent austerity,” 311–15; political resources of low-income people, 293; public opinion of poor, 310–11; research considerations, 292–95; social insurance programs, 298–302; taxbased transfer policies, 305–6 polling, 325–26 pollution, 84 Poole, K., 318 poverty: Americans’ attitudes toward, 310–11; children living in, 230, 264, 279, 300; children’s development and well-being impact, 119, 120; and educational attainment, 149; elite dissensus on causes of, 323; NHANES measure of, 86; parenting impact, 126; percentage of people living in, 228; “pro-poor” policy agendas, 326–27; racialized images of, 310–11; single mothers living in, 4, 230–32; and welfare reform, 2, 3. See also working poor poverty line, 4 poverty reduction. See distributional effects preschool children, 123, 277. See also early childhood education principal-agent problem, 163 prisons and prisoners. See incarceration private schools, 170 program evaluation, 72n58 Project STAR, 155, 156, 159, 167 property crime, 213, 214 PRWORA (Personal Responsibility and Work Opportunity Reconciliation Act) (1996), 1–2, 121, 216, 310–11 public finance. See government funding public opinion and support, 309–11, 324–26 Public Use Microdata Samples (PUMS), 189, 192 racialization, 310–11 Rainwater, L., 264, 273

/

345

Index random-assignment studies, 120–22, 130 Raphael, S., 199, 206, 217, 221n4, 221n9 reading skills, 150, 151, 167 Reardon-Anderson, J., 255n3 recessions, 83–84, 292 records, criminal history, 186, 198–99, 217, 218–20 Reich, M., 67n11, 68n24 relatives, financial assistance from, 247 Republican Party, 305–6, 307, 308, 312, 315–18 research considerations: children’s well-being, 118; crossnational comparison, 261–62; disconnected single mothers, 228, 254; early childhood education, 17; health and employment, 79–85, 107–8; health insurance, 14, 108; labor-market policies, 25–27; political issues and constraints, 292–95; school reform, 149, 176; school-towork programs, 50–51; self-sufficiency, 4–6; working poor, 261–62, 263 research methodology. See methodology Reshaping the American Workforce in a Changing Economy (Holzer and Nightingale), 13 retirement, 81 risk-taking behaviors, 84 Rivkin, S., 156 Rockoff, J., 156 Romano, R., 172 Romich, J., 268 Rosenbaum, J., 59–60 Rosenthal, H., 318 Rothstein, D., 51, 53, 55, 70n43 Rothstein, J., 63, 164, 168 Rouse, C., 149, 154, 166, 167, 171 Ruhm, C., 79, 80, 82–83, 84 Sabia, J., 37 Sabol, W., 204 Saffer, H., 83 Sampson, R., 222n14 San Antonio, Tex., living-wage law, 39 San Diego, Calif., living-wage law, 39 San Francisco, Calif., minimum wage, 68n20 San Jose, Calif., living-wage law, 39

346

/

Santa Fe, N. Mex., minimum wage, 68n20 Sass, T., 173, 174 savings, 13–14 Scandinavian model, 269, 271–76, 278–79 Schanzenbach, D. Whitmore, 165 S-CHIP (State Child Health Insurance Program), 267, 296, 298, 307–8, 318 Schneider, M., 169 Schoeni, R., 247 Schofield, D., 82 Scholz, J., 43, 63, 71n51 school achievement: across childhood, 122–23; charter schools, 173–74; and child care use, 132–34, 136–37; and class size reductions, 155; employment-based welfare and antipoverty program impacts, 124–35, 136, 137–40; and family income, 120, 126–30, 134, 136–37, 152; measurement of, 122; and parent’s education, 150–51; and school accountability, 167; and school spending, 153–54; and teacher merit pay, 159–60; and Teach for America teachers, 161–62; and vouchers, 170–72 school choice, 167–75 school finance, 152–56 schooling. See educational attainment school meal programs, 166, 308 school readiness, 150, 277 school reform: accountability, 49–50, 163–67, 169–70, 176; class size reductions, 154–56, 176; finance, 152–56; marketbased policies, 162–63; No Child Left Behind, 49, 65, 163–64; research considerations, 149, 176; school choice, 167–75; teacher quality, 156–62, 175–76 School-to-Work Opportunities Act (STWOA), 49, 50, 72n55 school-to-work programs, 48–60, 61, 64–65 school vouchers, 168–72 Schweitzer, M., 30, 32, 36 screening, employment, 198–200, 219 Seefeldt, K., 232, 255n3 Seidman, L., 43

self-sufficiency, research overview, 1, 4–6 Self-Sufficiency Program (SSP), 125, 135, 141 sentencing policy, 206, 209–11, 215 SES (socioeconomic status), and educational attainment, 150 sex offenders, 220 Shaikh, A., 80 Shaikh-Vytlacil (SV) bounds, 101–3, 108–9 Shaper, A. Gerald, 81, 82 Shapiro, R., 325–26 Siegelman, P., 215 signing bonuses, 158 Simmelick, J., 268 single men, 63 single mothers, 62, 119, 230–32. See also disconnected single mothers single persons, EITC effects, 45, 62–63 SIPP (Survey of Income and Program Participation), 21n17, 233–36, 238, 240 skill acquisition, 41–43, 65–66. See also educational attainment; training programs Skocpol, T., 299 small businesses, 322 Smeeding, T., 247, 264, 273 Smith, J., 70n40 smoking, 82, 83–84 Snyder, S., 85 social connectedness, 87, 89, 92–93, 103 social insurance programs, 267, 295, 298–302. See also Unemployment Insurance (UI) Social Security, 85, 267, 299–300, 312, 328 Social Security Act (1935), 303 social spending. See government funding socioeconomic status (SES), and educational attainment, 150 Sokejima, A., 84 solely state-funded programs (SSFs), 251–52 Soss, J., 332n34 Sparer, M., 320 spatial mismatch, of housing and jobs, 13 special education, 153, 166 special interest groups, 302, 322 spending. See government funding

Index SSFs (solely state-funded programs), 251–52 SSI (Supplemental Security Income), 237, 243, 245, 250–51, 303 SSP (Self-Sufficiency Program), 125, 135, 141 Stagner, M., 255n3 standard of living, minimum level of, 5 State Child Health Insurance Program (S-CHIP), 267, 296, 298, 307–8, 318 states: EITC supplements, 69n28, 271; policy responsibility, 19, 318–22, 328; solely state-funded programs, 251–53 statin drugs, 94, 98 Steiner, G., 331n13 Stern, D., 50, 71n44 Stevens, A. Huff, 152 stigmatization, 47, 62, 198–200 Stoll, M., 199, 206, 217, 221n4, 221n9 Stone, D., 331n19 strategic disagreement, 327 stress, work-related, 84 student achievement. See school achievement Student-Teacher Achievement Ratio (Project STAR), 155, 156, 159, 167 student-teacher ratios, 154 STWOA (School-to-Work Opportunities Act), 49, 50, 72n55 subsidies, wage, 26, 46–48, 61 subsidized child care, 264, 267, 276–79, 280, 296, 298 substance abuse, 236, 252 suicide rates, 82 Sullivan, J., 231 Supplemental Security Income (SSI), 237, 243, 245, 250–51, 303 supply and demand, 25, 26, 61 Supreme Court, 294 Survey of Income and Program Participation (SIPP), 21n17, 233–36, 238, 240 SV (Shaikh-Vytlacil) bounds, 101–3, 108–9 Sweden: child and family allowances, 273; child care, 278; child poverty, 264; family leave, 275, 276; unemployment and mortality study, 81; Unemployment Insurance, 273 Sweeten, G., 203

TANF (Temporary Assistance for Needy Families). See Temporary Assistance for Needy Families (TANF) Targeted Jobs Tax Credit (TJTC), 46–47, 48 targeting, of benefits, 17, 299 tax-based transfer policies, 265, 271, 305–6, 315 tax credits: Child Tax Credit (CTC), 271, 306, 314, 318, 328; Dependent and Child Care Tax Credit (DCTC), 277; for employers, 26, 46–48; in U.K., 280; in U.S. vs. other OECD countries, 265. See also Earned Income Tax Credit (EITC) tax cuts, 312, 314 tax deductions, 271 tax revolts, 312 teachers: alternative pathways for, 160–62; class-size reduction policy impact, 155–56; compensation, 157–60, 175–76; New Hope program survey, 136; quality, 153–54, 156–57, 278–79 Teach for America (TFA), 161–62 Tech Prep, 53, 70n42 teenagers. See youth Temporary Assistance for Needy Families (TANF): creation, 304; federalism impact, 319–20; funding, 296, 298, 314, 329; maintenance-of-effort requirements, 251, 253; percentage of poor families receiving, 271; politics of, 304; proposed revisions, 252–53, 254; studies of, 121, 232; value of benefits, 271; Work Opportunity Tax Credit, 26; work requirement revisions (2006), 248–49 Tennessee, Student-Teacher Achievement Ratio (Project STAR), 155, 156, 159, 167 Teske, P., 169 test-based educational reforms, 49–50, 163–67 test scores, 159–60, 173 Texas, charter schools in, 173 TFA (Teach for America), 161–62 Theorell, T., 84 Tiehen, L., 246 time limits, 121, 140, 238, 253, 267–68, 304 time-series studies, 27

TJTC (Targeted Jobs Tax Credit), 46–47, 48 tobacco consumption, 82, 83–84 trade-offs, 268–69 training programs: and children’s school achievement, 130, 132; funding of, 12–13; minimum wage effects, 42; social returns, 70n40; success and potential of, 17; for teachers, 161–62; for youth, 49 Travis, J., 189 treatment effect, of working on health, 81, 98, 100–103, 105, 106–7 trucking industry, 219 Tulsa, Okla., IDA program study, 13–14 Turner, J. Blake, 85 Turner, L., 232, 255n3 Turner, M., 42 turnover, job, 3 UI (Unemployment Insurance). See Unemployment Insurance (UI) unemployment: generous benefit program impact, 284; health impact, 79, 80, 81–85; local rates, 100–101; and parenting behavior, 126; in Wisconsin, 20n8. See also disconnected single mothers Unemployment Assistance (UA), 272 Unemployment Insurance (UI): eligibility, 270–71; employment impact of expansion, 284; future of, 310; government spending on, 296, 298; incarceration and employment, 204–5; in Nordic countries, 272–73; politics of, 301 uninsured, 14, 98, 108, 309–10 unions, 322 United Kingdom: child poverty reduction, 279; social benefits package, 279–80; unemployment and health studies, 81, 82 universal programs: British child allowance and tax benefits, 280; Chilean school vouchers, 172; early childhood education, 285–86; family leave, 275–76, 285–86; health insurance, 301, 320–21; income adjustments, 282–83; Nordic child and family allowances, 273

/

347

Index unmarried mothers, 62, 119, 230–32. See also disconnected single mothers Urban Institute, National Survey of America’s Families, 14 Urquiola, M., 168 U.S. Census Bureau, 149, 189, 192, 205 U.S. Department of Education, 159 U.S. Department of Labor, 110n6 vending machines, in schools, 166 Vigdor, J., 158 violent crime, 212, 213, 214 Vistnes, J. Primoff, 83 vouchers, school, 168–72 Vytlacil, E., 80, 101–3, 108–9 wage floors, 25, 26. See also minimum wage wages. See earnings and wages wage subsidies, 26, 46–48, 61 Wagstaff, A., 110n3 Waldfogel, J., 204 Wascher, W., 28, 30, 32, 36, 42, 45, 46, 70n34 Washington, D.C.: minimum wage, 68n20; school vouchers, 170 Washington, employment consequences of incarceration, 204 wealth accumulation, 13–14 welfare benefits, 67n6, 245, 271, 302–4 welfare caseloads, 2, 249 welfare leavers, 3–4, 236–40. See also disconnected single mothers welfare migration effect, 319 welfare recipients, racialized images of, 310–11

348

/

welfare reform, 1–3, 228, 262, 304 well-being, of children. See children’s well-being Western, B., 192, 203 Wheaton, L., 254n2 Whiteford, P., 264, 273 Whitehall studies, 79 whites, incarceration rates, 185–86, 189, 190, 192 Whitmore, D., 155, 165 WIC (Women, Infants and Children), 308 William T. Grant Foundation, 71n46 Wilson, W. Julius, 299 Winicki, J., 166 Winkler, A., 232 WINTC (Work Incentives Tax Credit), 46 Wisconsin: earnings of welfare leavers, 3–4; unemployment, 20n8; welfare benefit reductions with increased earnings, 268; welfare reform, 20n8. See also Milwaukee, Wis. women: health status, 91–93, 95–97, 99, 100, 102–3; laborforce participation, 1, 15–16; school-to-work participation effects, 56, 57, 58–59; uninsured, 14 Women, Infants and Children (WIC), 308 work-based welfare reform, research considerations, 1, 3–4. See also specific index headings work experience, 197–98 work hours, 84, 108 Work Incentives Tax Credit (WINTC), 46 working poor: cost of increased benefits for, 284–86; family leave, 263, 274–76, 280;

health, 106, 107–8; income benefits, 269–74; policy considerations, 107–8, 268–69; research considerations, 261–62, 263; social benefits package structure issues and recommendations, 264–68, 280–86; statistics, 262; subsidized child care, 264, 267, 276–79, 280 Work Opportunity Tax Credit (WOTC), 26, 46, 48 workplace stressors, 84 work-release programs, 204–5 work requirements, under AFDC/TANF, 248–49, 251–53, 304 work support programs: child care, 132, 134, 136–37; children’s school achievement impacts, 122–26, 140; costbenefit analysis, 135–36; data sources and methodology, 120–22; earnings supplements, 121, 124, 129–30, 135–36, 139–40; solely statefunded programs (SSFs), 251–52. See also Earned Income Tax Credit (EITC) World Bank, 160 WOTC (Work Opportunity Tax Credit), 26, 46, 48 Wu, X., 38, 45, 67n19, 69n34 youth: high-risk youth policies, 205–6; minimum wage effects, 27–28, 29, 42–43; poverty and school achievement, 120; school-to-work programs, 48–60, 61, 64–65 Zedlewski, S., 267 Zhu, P., 170 Ziliak, J., 38, 68n19, 69n34 Zimmerman, S., 267