The nation's social welfare policy reflects an ongoing effort to balance sometimes

Transcription

1 Journal of Economic Perspectives Volume 4, Number 1 Winter 1990 Pages Work and Welfare: Lessons on Employment Programs Judith M. Gueron The nation's social welfare policy reflects an ongoing effort to balance sometimes competing objectives alleviating poverty and promoting self-sufficiency in a manner consistent with underlying public values about the primacy of the family and the importance of work. Concern has been growing that the welfare system has not been doing this very well, and welfare reform once again moved towards the top of the policy agenda, resulting in passage of the Family Support Act of 1988 (FSA). That the bill was passed is evidence of broad agreement about elements of the problem and its solution. That it was passed amidst acrimony and with relatively modest provisions reflects continuing disagreement about the fundamental objectives of reform. This paper discusses what economists know about the potential of one central component of the new legislation: the effort to transform welfare from a means-tested entitlement into a reciprocal obligation, in which getting a welfare check would carry with it a requirement to look for and accept a job, or to participate in activities that prepare people for work. It sets the context for this discussion by briefly outlining why this approach to reform gained support and by summarizing major policy and program alternatives. Why Welfare Reform? The extent of poverty in the United States and the choices facing poor people are affected by a broad range of policies that influence the distribution of income and the Judith M. Gueron is President of Manpower Demonstration Research Corporation, New York, New York.

2 80 Journal of Economic Perspectives rewards of work. Public assistance (welfare) programs the most visible of which is Aid to Families with Dependent Children (AFDC) are means-tested: that is, eligibility and benefits depend on income and assets. Other policy tools include social insurance programs (such as Social Security or unemployment insurance), which are available to everyone and are based on age, disability, or employment, and a range of "nonwelfare" programs including child support enforcement and tax and education policies. Welfare has few supporters and many critics, yet it has resisted repeated efforts at major reform. The fundamental dilemma is the impossibility of simultaneously maximizing the two primary policy objectives: reducing poverty and encouraging self-support. Simple logic dictates a trade-off, at least in the short run and probably in the long run. Because welfare programs are means-tested, they discourage work, since the more you work, the less assistance you receive. Moreover, any effort to increase benefits to combat poverty more effectively will only further decrease the incentives for recipients to take low-paying jobs and work. The concern that generous public assistance would undermine self-reliance has a long history. It was central to the debates about the seventeenth-century English poor laws, which established the principle that the state is responsible for aiding the poor, and it has been voiced repeatedly in this country by public figures ranging from Benjamin Franklin to Franklin D. Roosevelt, Lyndon Johnson, and Ronald Reagan. The dilemma could be resolved if the poor could be divided into two groups those who are able and expected to work and those who are not and assistance limited to the second group. While the vocabulary has changed with the century, the attempt to categorize the poor has been fundamental to the rhetoric and design of assistance programs. The Social Security Act of 1935 (SSA), which provided the first federal support for welfare programs, adopted this approach and limited assistance to the aged, the blind, and dependent children in single-parent households. (SSA created the Aid to Dependent Children ADC program, which was later retitled AFDC when benefits were expanded to cover mothers as well as children.) This categorization has worked smoothly for the aged and severely disabled, who have been provided with increasingly generous assistance with relatively little concern about its impact on work incentives. The continued controversy has centered on single mothers, the primary adult beneficiaries of AFDC. Expectations for single mothers have undergone major shifts in the past century. The pre-twentieth-century view that such women should work was slowly changed in the early 1900s. This was embodied in the Social Security Act, which represented a "national commitment to the idea that a mother's place is in the home" (Steiner, 1971, p. 54). 1 Designers of the AFDC program hoped that it would provide (what was then thought to be) a relatively small group of poor widows with 1 See Garfinkel and McLanahan (1986) and Katz (1986) for recent discussions of the evolution of policy on the employability of women.

3 Judith M. Gueron 81 sufficient income to stay home and care for their children in accordance with prevailing middle-class norms. Concern about work incentives was muted. Developments over the past 30 years, however, have undermined support for the 1935 vision of the AFDC program. First, employment rates for all women, including single parents and women with very young children, have increased dramatically. Thus, 65 percent of all women with children and 57 percent of married women with children under the age of 6 participated in the labor force in 1987 compared to 30 percent and 19 percent, respectively, in 1960 (Reischauer, 1989). With the majority of mothers working at least part time and often from economic necessity, it became difficult to defend the equity of excusing single mothers who were rearing children from the obligation to work. The argument suffered too from the absence of clear research evidence of a link between a mother's employment and the well-being of her children. (See, for example, Garfinkel and McLanahan, 1986.) Second, in contrast to original expectations, AFDC caseloads and costs grew rapidly in the 1960s and early 1970s. Although caseloads later stabilized, and current costs account for less than one percent of the federal budget, the perception of high cost remains widespread. In fact, the more recent increases in costs have been for in-kind programs, particularly food stamps and Medicaid. 2 The nature of the caseload also changed: currently, more than 90 percent of all women heading AFDC cases have never been married or are divorced or separated rather than widowed. Historically, the welfare system had discriminated among technically eligible groups in its behavioral requirements and administrative practices, and this shift in caseload subtly and not so subtly affected public support. Third, in what Senator Daniel P. Moynihan refers to as "the earthquake that shuddered through the American family," there has been a dramatic increase in the number of female-headed families, the major demographic group most at risk of poverty. For example, in 1987, 21 percent of all children lived with single mothers compared to 7 percent in Even more dramatically, more than half the children born today will spend some time in a mother-only family and half of all such families are likely to be poor (Garfinkel and McLanahan, 1986, p. 1). Concern about the resulting high rate of child poverty has focused attention on the minimal support by absent fathers, the limited capacity of women on welfare to work their way out of poverty, and the very design of the AFDC program, which in half the states has been available only to single-parent households. In response, Congress has strengthened child support enforcement and (since 1988) required states to extend benefits to two-parent families in which the principal earner is unemployed. While creation of this relatively small AFDC-UP (AFDC-Unemployed Parent) program increased incentives for families to stay together, it raised anew the question of work behavior, since it extended benefits to a clearly employable group. Fourth, although recent research confirms that most people use welfare for short-term support only, it also points to a substantial minority who remain poor and dependent for long periods: Approximately one-quarter of those who ever use AFDC 2 See the accompanying paper by Gary Burtless.

4 82 Journal of Economic Perspectives stay on welfare for ten or more years. They account for 60 percent of those on the rolls at any one time and at least 60 percent of the cost of AFDC (Ellwood, 1986, p. xii). Such findings have intensified concern about the high cost of supporting these families and the possible negative effects of welfare on both the mothers and their children. While it is not clear to what extent the design of the AFDC program has contributed to these developments, they have focused attention on whether welfare programs appropriately address the fundamental tension between the competing values of self-reliance and compassion. The four changes described above undermined the idea that welfare should provide an alternative to work and spawned resistance to using AFDC benefit increases to combat poverty. (AFDC benefits were allowed to decline by about 25 percent in constant dollars between 1975 and 1985.) Instead, policy has increasingly emphasized the responsibility of parents, mothers as well as fathers, to support their children. But questions are also being raised about the structure of AFDC, and the strong work disincentives created by its high implicit tax rates and the linkage between AFDC benefits and the entitlement to Medicaid health insurance. The challenge is how to reshape social policy to emphasize parental responsibility without increasing the risk that children of parents who fail to fulfill their responsibilities will suffer. 3 Reform Options Over the years, debate has focused on two broad approaches, currently referred to as "welfare" and "nonwelfare" reform, which place different emphases on the goals of reducing poverty and dependence. "Welfare" reformers have focused on changing the AFDC program either to create more effective incentives to reduce dependence, or to devise requirements and provide services that directly encourage employment. Advocates of the second, "nonwelfare" approach stress that reforms within the welfare system are likely to be inadequate. In general, they endorse more far-reaching, and often substantially more expensive, reforms designed to assure that the kinds of jobs realistically open to welfare recipients will provide income that is above the poverty level, for example, through raising the minimum wage and the Earned Income Tax Credit (EITC). 4 Welfare Reform The welfare reform strategy seeks to encourage self-support by changing the expectations and incentives within the AFDC program. This strategy took many forms 3 Garfinkel and McLanahan (1986) call the "new American dilemma" the government's choice between reducing poverty among women and children by offering more generous benefits or reducing the prevalence and dependency of such families by cutting benefits. 4 The Earned Income Tax Credit is a tax supplement to earnings for taxpayers who maintain a household for one or more children. If a worker's family income falls below a certain level, the federal government funnels money to him or her by cutting his or her taxes. The credit is also "refundable": If a family pays no federal taxes, it stills gets the benefit in the form of a check from the government.

5 Work and Welfare: Lessons on Employment Programs 83 in the 1960s. Some approaches emphasized rehabilitation through social services. Others favored modest financial incentives (by changing the rules for determining welfare benefits) in the hope that recipients would voluntarily choose to work, even if that reduced their welfare benefits. Neither social services nor the increase in financial incentives seemed to have much impact on caseloads or on the share of recipients who worked. Moreover, research pointed to the high cost of the incentive approach: Cutting welfare tax rates could actually expand the size of the welfare population and even lead to an aggregate decline in work effort. While these findings contributed to a return to higher AFDC tax rates in 1981, restiveness with the resulting obviously high work disincentives, especially when combined with income-conditioned health insurance and child care costs, have led to some restoration of AFDC work incentives and a movement to provide at least transitional in-kind benefits to people who work their way off welfare. 5 Given the disillusionment with incentives and social services, attention shifted starting with legislation in 1971 towards requirements and the provision of employment-related services. While the role of carrots and sticks varied, the basic thrust was to move AFDC from an income-conditioned entitlement towards a "reciprocal obligation" or "new social contract." Under this approach, the welfare recipient is required to look for and accept a job or to participate in work experience, education, or training to prepare for work, or risk losing some welfare benefits. The government is responsible for providing, in addition to the basic grant, the services and supports that will help the recipient obtain and retain employment. In this paper, the phrase "welfare employment programs" refers to programs that attempt to implement this concept. (While some have argued that these requirements will transform AFDC into a transitional or short-term program, actual legislated changes have been less stringent: People satisfying the program's participation requirements continue to receive full grants of indefinite duration; those not satisfying program requirements usually continue to receive partial grants.) This approach, initially advocated by conservatives, now receives broader support, as seen in proposals by the nation's governors that preceded the passage of the Family Support Act (National Governors' Association, 1987) and in the legislation itself, as discussed below. However, the rhetoric of support should not mask the extent to which persistent differences in objectives or values and in views of social justice or the causes of poverty and unemployment translate this broad agreement into highly diverse specific programs. To simplify a complex debate, in which there is more of a continuum than a dichotomy of views, liberals and conservatives stress different designs and outcomes. Liberals generally accept the premise that work is better than welfare but continue to emphasize reducing poverty over reducing dependence. They argue that welfare recipients want to work but lack the education and skills to get jobs that assure 5 The accompanying paper by Gary Burtless summarizes the numerous changes in AFDC work incentives enacted over the past 20 years, the way in which benefit-reduction rates combine across transfer programs to limit the income gained from work, and the results of the earlier Negative Income Tax experiments.

6 84 Journal of Economic Perspectives self-sufficiency and a decent standard of living, and that many of them would not be able to find adequate care for their children if they did find jobs. Thus, liberals prefer programs that serve volunteers first, offer choices, provide intensive education and training services, do not require people to take low-wage jobs or indeed any job, and assure adequate child care and health insurance to those who leave welfare to work. Liberals focus on insufficient human capital as the critical reason for continued dependence, and they believe that countering this will have high returns. Conservatives, on the other hand, usually emphasize reducing dependence over reducing poverty. They argue that jobs are available and believe that welfare recipients are unwilling to work, are too discouraged to try, or have unrealistic expectations about their job prospects. Thus, conservatives favor programs that set clear expectations for recipients, require participation in program activities or regular employment, provide low-cost job placement assistance rather than expensive training, and mandate "workfare" for those who remain on the rolls. 6 They tend to doubt that providing recipients with more human capital will enable them to get jobs that meet their unrealistically high expectations, or that such efforts can be cost-effective. In sum, liberals look to increases in earnings and reductions in poverty, while conservatives stress welfare savings and the intrinsic value of program participation and work performed while on welfare. The Family Support Act of 1988 (FSA), especially its centerpiece, the Job Opportunities and Basic Skills Training (JOBS) Program, redefines federal policy on welfare employment programs and represents a compromise among these conflicting visions of welfare reform. The key elements of FSA are that parents, both fathers and mothers, should be the primary supporters of their children and that government should provide incentives and assistance to welfare recipients to find employment. For absent fathers, this translates into greater enforcement of child support collections and new obligations on welfare mothers to cooperate with such efforts. For welfare recipients, JOBS contains features of both the conservative and liberal views: an expansion of the program participation mandate to women with younger children (as young as age 3, or age 1 at state option), a complex balance of mandatory and voluntary elements (including an emphasis on serving volunteers first and on determining services through case management and client choice), child care guarantees, and investments designed to improve the capacity of AFDC mothers to obtain jobs and, hopefully, self-sufficiency. While JOBS gives states substantial flexibility in designing welfare employment programs, it includes incentives to move the system in new directions. In contrast to state programs of the 1980s, which typically provided job search activities followed by unpaid work experience, JOBS places much greater emphasis on education. It requires that (to the extent resources are available) education be offered to any adult on AFDC who lacks a high school diploma or does not demonstrate basic literacy. 6 Throughout this paper, "workfare" refers to a mandatory work-for-benefits program using either the Community Work Experience Program or WIN (Work Incentive) work experience approaches and not the evolving broader definition that encompasses any work-related activity.

7 Judith M. Gueron 85 States must also provide job skills training, job placement services, and two of the following: group or individual assistance in locating a job (job search), on-the-job training, or community work experience (workfare). JOBS also introduces a new vision for teenage custodial parents by requiring that they participate in an educational activity (the "learnfare" provision) and adds a stronger work mandate for AFDC-UP adults. Other provisions of FSA authorize funding for in-program child care for JOBS participants, require states to provide benefits to unemployed parents, and provide for a year of transitional child care and Medicaid for individuals leaving welfare for employment. Nonwelfare Reform Proponents of nonwelfare reform argue that poor women are unlikely to move from welfare to above-poverty-level incomes, at least in the short run and even after participating in an employment or education program, that reducing poverty should be the fundamental goal of reform, and that policies should equitably treat the working and the dependent poor. Maintaining that AFDC and other income-conditioned programs are inherently flawed because they cannot "reflect and reinforce our basic values" and are also stigmatizing and degrading (Ellwood, 1988, p. 237), this group seeks to institute policies outside the means-tested welfare system that increase the returns from part- or full-time work and provide alternative sources of income. Examples include strengthening the child support system (or even guaranteeing a child support payment), "making work pay" by either increasing the minimum wage and the Earned Income Tax Credit or requiring employers to provide health insurance, improving the education and training systems, and providing jobs directly if there is an insufficient demand for labor. Advocates of nonwelfare reform stress parental responsibility but seek specific measures to assure that work will result in adequate income and security. This greater generosity towards the working poor is often combined with restrictions on the ability of those not working to get long-term welfare, or replacing AFDC with a transitional program focused on employment and jobs. This reorientation would make work more rewarding, but would offer no increase and possibly even a reduction in support for those who do not work. 7 In theory, welfare and nonwelfare reform are complements, since work mandates should be more effective in an environment in which work pays. In practice, budget constraints may make them alternatives. With the passage of FSA, which contained elements of both approaches (in the JOBS title and in strengthened child support enforcement), Congressional action in 1989 shifted towards reforms that would assist the non-dependent poor. The remainder of this paper summarizes what is known about the effectiveness of welfare employment programs. At the end, it returns to a discussion of the tradeoffs between welfare and nonwelfare reform. 7 Ellwood (1988), Garfinkel and McLanahan (1986), and the Ford Foundation (1989) outline specific nonwelfare reform policies. Reischauer (1989) discusses the constraints on efforts to reform welfare.

8 86 Journal of Economic Perspectives Reforming Welfare with Work The current push for work requirements should be viewed as only the latest version of an ancient alternative with a very checkered history. 8 What is new, however, is that the current discussion draws less on assertions of competing values and more on information about what can realistically be accomplished, based on studies of state experiences during the early and mid-1980s. State welfare programs for able-bodied men have often included work requirements. Congress first expressed a preference for "reciprocal obligations" in the AFDC program in 1971, when it required all adult recipients with school-age children to register and participate in a welfare employment program the recently enacted Work Incentive (WIN) program and take jobs, or risk grant reductions. Because of meager appropriations, however, participation was often limited to registration, and the program lost credibility as it failed both to meet its operational objectives and to generate reliable information on caseload changes and cost-effectiveness. Several state-run demonstrations in the 1970s also generated more questions than answers as attempts at welfare reform foundered against bureaucratic resistance, legal challenges, and implementation problems. 9 While the large-scale feasibility and cost-effectiveness of mandatory work programs remained uncertain, some careful social experiments pointed to the realistic potential of smaller-scale voluntary employment strategies. The National Supported Work Demonstration (implemented in 15 sites between 1975 and 1980) showed that structured, transitional, paid work experience could have positive long-term effects for very disadvantaged welfare recipients and could also be cost-effective for taxpayers, despite high initial costs. For example, by months 19 through 27 after enrollment, a period during which most people had been out of the program for close to a year, the supported work group had average monthly earnings that were $77 (more than 45 percent) higher than those of women in the experiment's control group, and average monthly AFDC payments that were $52 (23 percent) less than those of controls (Manpower Demonstration Research Corporation, 1980, p. 153). Evaluations of programs that assisted women in searching for jobs indicated that this low-cost approach could also provide lasting benefits to a wide range of recipients (Wolfhagen with Goldman, 1983). Both programs showed clear impacts, although not large ones, and indicated greater effects for the more disadvantaged. While encouraging, these studies did not address the question of whether largescale mandatory programs could have similar effects. In the Omnibus Budget Reconciliation Act of 1981, Congress gave states an option to convert WIN into a block grant administered by welfare agencies and to use workfare for the first time as a component of state programs. This action gave states the flexibility to put their own 8 Katz (1986) provides a useful summary of the continuity between the current debate and earlier controversies, including the debate about the relative merits of "outdoor" relief (cash assistance) and "indoor" relief (in a poorhouse that required one to work). 9 See Mead (1986) and Rein (1982) for different views of the WIN program, and see Gueron and Nathan (1985) for a summary of pre-1980 state demonstrations.

9 Work and Welfare: Lessons on Employment Programs 87 stamp on what had been seen as a federally mandated and bureaucratic program, but less money to do it with. WIN, the major federal funding source for these programs, had become a target for annual budget cuts, with funding falling 70 percent between 1981 and 1987 in nominal dollars. The combination of increased flexibility and reduced resources led to a mixed response. Overall, the scale of work programs declined, with fewer people and areas of the country covered. But what remained, even if usually modest in design, seemed reenergized, as states reorganized staff and shaped work initiatives to reflect prevailing values and experience. Studies by the U.S. General Accounting Office (1987) and the Urban Institute (Nightingale and Burbridge, 1987) show that in 1986, 37 states were operating some form of job search assistance or workfare, the most common activities; however, these were implemented in areas covering only about 40 and 30 percent, respectively, of the national AFDC caseload. But because of the absence of reliable and consistent state data, these reports offer no firm estimates of the percentage of the caseload participating in these programs, nor of more complex outcomes such as the programs' cost-effectiveness or impact on employment or welfare dependence. Much of what is known about these questions comes from the smaller number of states with welfare employment programs that were carefully evaluated using largescale controlled social experiments involving almost 40,000 people. Recent work confirms the importance of this approach in determining the effectiveness of social programs. 10 Extensive research has shown that, contrary to popular belief, many recipients move off the welfare rolls fairly rapidly. To find out if an employment program changes earnings or welfare dependence, it is important to be able to distinguish between program-induced changes and the normal dynamics of welfare turnover. While many approaches have tried to obtain such estimates, including those using time series studies and comparison groups, the most accurate employ the techniques of social experimentation, in which welfare applicants or recipients are randomly assigned either to the mandatory work program or to a control group excused from program requirements. The behavior of these two groups can then be tracked over time, with differences between them providing estimates of program impacts. The seven social experiments summarized here are of programs generally representative of mandatory state welfare employment initiatives during the early and mid-1980s, and include projects in Baltimore, Cook County (including Chicago), and San Diego (two different approaches), as well as multi-county efforts in Arkansas, Virginia, and West Virginia. These studies answer several questions important to the debate on welfare employment strategies. Further detail on these programs is available in Gueron (1987) and reports by the Manpower Demonstration Research Corporation (MDRC). 10 See Burtless and Orr (1986), Ashenfelter (1987), and Lalonde and Maynard (1987) for discussions of the relative merits of experimental and nonexperimental studies.

10 88 Journal of Economic Perspectives Program Design and Participation The participation requirements that these state initiatives imposed varied along a number of dimensions: the duration of the obligation, the activities provided, who was required to participate, and the extent to which the requirement was enforced. Table 1 summarizes the seven programs along these dimensions. Duration of Obligation. While all seven programs were mandatory that is, people could lose some benefits if they refused to meet program requirements five imposed obligations that usually lasted no more than four months. Only two, West Virginia and the San Diego Saturation program, mandated participation as long as Table 1 Dimensions of welfare employment programs in seven areas

11 Judith M. Gueron 89 someone remained on welfare. In fact, the San Diego Saturation program was a federally funded demonstration designed to test the feasibility and impact of imposing an open-ended participation requirement. Activities Required. In general, the states offered limited services at relatively low cost, ranging from under $200 per eligible person in Cook County and Arkansas to around $1,000 in Baltimore. Except for West Virginia, which required recipients to work in exchange for their benefits, the states chose to implement a participation rather than a strict work requirement. Typically, welfare applicants or recipients were required to look for a job with program assistance for two to four weeks. In a two-week version, the first week might consist of three-hour-per-day group sessions designed to build self-confidence and job-seeking skills; the second week might consist of two to three hours per day of supervised telephone contact with prospective employers. If not successful, program registrants might be required to work for up to three months doing unpaid work (workfare). Usually, these were entry-level jobs in public or nonprofit agencies involving maintenance, clerical, park upkeep, or human services functions. Monthly work hours equaled the welfare grant divided by the minimum wage. After this, if they were still not employed, the obligation would end or some minimal job search would periodically be imposed. This was the pattern in Arkansas, Cook County, and Virginia, and in San Diego's first project, although the programs differed substantially in the extent to which they provided assistance in job search and actually enforced participation in workfare. San Diego's Saturation program mandated education or training for those who completed the sequence without finding a job and were still on welfare. In contrast, the Baltimore program did not impose a fixed sequence of activities but tried to offer a choice of job search, unpaid work experience, or a range of short- or longer-term educational and training activities. Mandatory Participants. Severe budget constraints combined with concern about requiring women with young children to work led all of the states to limit the share of the caseload subject to the mandate. Except for Arkansas, which included women with children age 3 or over, the states imposed an obligation on only the one-third of the AFDC caseload with school-age children, the group that during this period was subject to the WIN mandate. (As noted earlier, FSA extends this to women with younger children.) Some states required all adults in this group to participate (spending very little per person), but others focused the program on certain areas of the state or certain subgroups. 11 Extent of Enforcement. Someone unfamiliar with welfare employment programs or with the daily situations facing many women on public assistance might expect participation in well-administered programs to be close to 100 percent. However, given substantial welfare turnover, temporary deferrals (because of part-time employment, illness, child care, or other reasons acceptable to the program), and normal administrative delays, this is clearly not reasonable. In fact, administrators during this 11 As indicated in Table 1, some states also required participation of parents in AFDC-UP cases. The discussion in this paper focuses on the results for single-parent AFDC cases.

12 90 Journal of Economic Perspectives period faced substantial uncertainty about the feasibility of requiring participation or the level of activity that would reflect the successful implementation of a participation mandate. In the seven programs, participation rates, however defined, were less than universal. In the five areas with short-term requirements, a useful way to measure participation was to track newly enrolled AFDC applicants or recipients and ask what fraction were ever active in the program or otherwise satisfied its requirements. Using this undemanding measure, typically about half the women had taken part in some activity within nine months of registration, and between 75 and 90 percent (depending on the location) had in some sense satisfied the program mandate (if not through participation, then by getting a job, leaving the rolls, or having their welfare grant reduced). (The rate of sanctions or temporary grant reductions varied from between 2 and 12 percent across the seven programs.) In areas with ongoing requirements, a more appropriate indicator is the share of the eligible caseload participating each month in the program or an alternative activity (such as employment or education) that satisfies the program mandate. The best data on this come from the San Diego Saturation program, which implemented a sequence of job search assistance, workfare, and referrals to community education and training programs. During a typical month, 52 percent of the WIN-mandatory caseload subject to the requirement participated in some relevant activity. Importantly, the monthly rate was very sensitive to the types of activities counted: if only the three sequenced activities were included, 33 percent of the caseload participated; if the 19 percent of those registered for the program who were employed in any month were counted, the rate rose to 52 percent. While these monthly rates may appear low, a number of factors suggest that 50 to 60 percent participation, broadly defined, probably represented the upper bound of administrative feasibility. Ninety percent of the registrants in the San Diego Saturation program who were eligible were either active in program services or inactive but had complied with program requirements during the month, even if they did not participate. Many of those inactive were assigned to components scheduled to begin during the next month; some were temporarily excused from participation owing to illness or other situational factors (Hamilton, 1988). Participation rates are also affected by the program model and local conditions. San Diego's relatively healthy economy during the period the Saturation program operated, its extensive network of education and training facilities, and California's high AFDC grant levels (which enabled more people to combine unsubsidized employment with the receipt of welfare than would be the case in other areas) suggest that other communities would have substantial difficulty replicating these rates. While the seven initiatives differed in terms of the liberal and conservative preferences for the design of welfare employment programs discussed earlier, in general they came out somewhere between the two. Primarily as a result of funding constraints, services were usually low-cost, offered little choice, and emphasized job placement rather than skills development. However, obligations were generally brief, sanctions were infrequent, and workfare did not emerge as the predominant activity.

13 Work and Welfare: Lessons on Employment Programs 91 Overall, the initiatives were incremental, not radical. As a result, questions remain about the feasibility and effectiveness of more intensive services or tougher work requirements, such as mandating more hours per week or more months of participation, stressing workfare rather than alternatives, or achieving higher participation rates among those eligible for the programs. Program Impacts and Cost-Effectiveness The results from the seven studies, summarized in Table 2, show that in most cases programs led to consistent and measurable increases in employment and earnings. 12 These positive impacts appeared to continue for at least three years in the states where data were available, with earnings gains of approximately 10 to 30 percent per eligible person in the caseload in the final year of the study. In Arkansas, for example, members of the experimental group were more likely to be employed, with increases of 4 to 6 percentage points in each of the three years of follow-up. By the third year, this translated into earnings gains of $337, or 31 percent, when experimentals as a group (including those who did not work as well as those who did) averaged $1,422 in earnings compared to $1,085 for controls. In Baltimore, increases in employment rates were smaller, but third-year earnings gains were larger, $511. Both the San Diego programs, serving two different groups of the WIN-mandatory population, also had notable impacts. 13 In general, these programs also led to welfare savings, although these were usually smaller and less consistent than the earnings gains. In Arkansas and the San Diego Saturation program, for example, welfare recipiency was reduced significantly (by 7 percentage points by the end of the last year of follow-up), with annual savings of $168 per person in Arkansas and $553 in San Diego (including those on the rolls as well as those off welfare) in the final year. In other states, there were smaller or no reductions in the rolls and variations in the effect on average welfare payments. However, the findings from two states provide important exceptions. First, the workfare program in West Virginia did not lead to increased employment and earnings. As program planners foresaw, this was probably because West Virginia is a rural state with exceptionally high unemployment. As a result, while the program could reinforce community preferences for work, keep job skills from deteriorating, and provide useful public services, it could not translate these into gains in unsubsidized employment. Second, the program in Cook County (which has since been ended) resulted in no statistically significant increases in employment and earnings, although there were small welfare savings. Here, the key explanation may lie in the program design. The Cook County program was the least expensive of those studied; mainly, it monitored people and sanctioned those who did not participate, providing little direct assistance 12 Throughout this paper, increases and decreases are noted only for changes that are statistically significant in Table 2 or in more detailed quarterly results. 13 These employment and earnings impacts and the benefit-cost findings discussed below assume that the increased employment of welfare recipients did not come at the expense of nonrecipients. The extent of any such displacement is unknown and very difficult to measure.

14 92 Journal of Economic Perspectives Table 2 Summary of the impacts on AFDC eligibles of welfare employment programs in seven areas

15 Table 2 (Continued) Judith M. Gueron 93

16 94 Journal of Economic Perspectives even in its job search component. These exceptions provide useful reminders of the importance of labor market conditions and possibly of the need to provide at least minimal assistance to get employment results (if not welfare savings). An analysis of five of the programs (Friedlander, 1988) showed, moreover, that gains were not distributed uniformly across the eligible caseload. The most employable people for example, women who were first-time welfare applicants and had recent work experience usually did not gain from program participation. While this group showed high program outcomes (as measured, for example, by employment rates), similarly high rates were found for comparable people in the control group. That is, even without special assistance, many of these women stayed on welfare only for relatively brief periods. In contrast, women with little or no recent work history benefited more consistently from these programs, even though the programs offered only limited assistance and almost no intensive training. While more of these women remained on welfare after participating in a work program, their performance relative to similar people in the control group was more impressive. However, Friedlander (1988) also identified a third, more disadvantaged group long-term welfare recipients with no recent employment who did not show consistent earnings impacts, despite some welfare savings. The JOBS legislation's focus on serving long-term recipients suggests the importance of determining whether more intensive services (including education and training) will be more successful in increasing this group's earnings and self-sufficiency. A detailed benefit-cost study examined the effects of these programs from several perspectives: the government budget, the welfare population, and society as a whole. In general, the results showed that the budgetary cost was more than offset by projected savings within two to five years of program enrollment and that, for the women on welfare, projected earnings gains usually exceeded estimated reductions in welfare benefits and other transfer payments (such as Medicaid and food stamps). The programs also scored positively from society's perspective, leading to gains in total social resources. Overall, these studies answer some key questions about the "welfare reform" option. First, with resources and time, such programs can be successfully implemented and can impose obligations on some share of the caseload. Second, even the relatively modest initiatives implemented in the early to mid-1980s led to a notable substitution of earnings for welfare and proved cost-effective, suggesting that the success can be repeated on a larger scale. The earnings gains are particularly important because they are averages that include all persons on welfare eligible for the program, not just the approximately 50 percent who actually participated or the people who found jobs. 14 Of course, average increases of about $300 to $650 a year conceal wide variations. For the people who benefit, through new jobs or higher earnings, the gains are much larger. 14 Both the impact and benefit-cost results were estimated for all persons eligible for the program, not just the 50 percent who participated. This is appropriate because mandatory programs spend resources processing nonparticipants and can have impacts on them if people change their behavior to avoid program participation or if welfare grants are reduced as a result of sanctions.

17 Work and Welfare: Lessons on Employment Programs 95 But the results also suggest caution in what can be expected from this type of reform: Alone, these programs do not offer an immediate cure for poverty or dependence. Their impacts are modest; many people remain dependent; and those who move off welfare often remain poor. Where Do We Go from Here? This paper has outlined two current approaches to restructuring social welfare policy: an expansion of welfare employment programs ("welfare reform") or the use of other tools to attack poverty ("nonwelfare reform"). The studies of the past seven years provide substantial information on the potential of the first approach, but the limited nature of the state initiatives described above suggests that two key questions remain unanswered. First, would either tougher work requirements (as advocated by conservatives) or more expensive education and training services (as advocated by liberals) make these programs more effective in reducing poverty or dependence? Different judgments on the answer to this question underlie the dispute that almost torpedoed the Family Support Act. Substituting reliable findings for rhetoric is particularly important because the new resources provided by the act's JOBS title will not be sufficient for states to offer comprehensive and intensive employment services to the full AFDC caseload. Instead, states will have to make choices. The JOBS legislation assures continued state flexibility, since it does not establish a uniform program but, rather, uses the carrot of higher federal matching rates as an incentive for states to satisfy certain design features: importantly, meeting targets for participation rates and for service to potentially long-term welfare recipients. 15 States moving to implement the JOBS legislation before the October 1990 deadline will be starting from very different positions and are likely to face widely divergent budget constraints, since overall JOBS resources will depend on the extent to which state legislatures put up money to draw down matching federal funds. As a result of the different pressures in JOBS, states will be called on to balance coverage and intensity, that is, using new resources either to extend lower-cost services to more people (for example, to meet participation targets) or to provide more enriched services to fewer people, or some combination of the two. The research to date does not provide clear guidance on whether greater investments in education and training will result in sufficiently greater returns to justify the additional outlays. While some studies of small-scale, voluntary programs suggest that more expensive services have higher returns, there is no evidence that 15 Thus, in order to avoid a reduction in federal matching rates, states must have at least 20 percent of the newly expanded mandatory caseload participating in program activities (not counting employment) in each month by 1995 (with higher rates for adults in AFDC-UP cases) and spend at least 55 percent of JOBS funds on families in which the custodial parent is under age 24 and has no high school diploma (or equivalency) or has had little or no work in the last year; the youngest child is within two years of ineligibility for AFDC; or the family received AFDC in 36 of the previous 60 months.

18 96 Journal of Economic Perspectives these can be replicated for a larger and more diverse population. (The San Diego Saturation program provides the first indication that this may be the case.) A second unanswered question is whether the results summarized in this paper can be replicated or improved upon if programs are expanded to serve a larger share of the caseload, particularly women with younger children. The new legislation places a priority on serving young mothers, based on extensive evidence of the high personal and social costs of their likely long-term dependence (Bane and Ellwood, 1983; Burt, 1986). Unfortunately, while there is a widespread consensus on the urgency of addressing the problems caused by teenage pregnancy, there is much less information on successful ways to assist young mothers and their children. Relatively little is known about what services or requirements will prove effective for this population, or about the adequacy or cost of providing needed child care. In particular, it is not clear whether low-cost mandates will be effective or feasible for this group, or whether very intensive services will be required to encourage and enable these young women to choose employment over long-term dependence. Current and planned state initiatives address both questions. For example, California's Greater Avenues for Independence (GAIN) and New Jersey's Realizing Economic Achievement (REACH) programs two of the more complex state JOBSfunded programs are designed to provide a much more extensive range of services and, in GAIN, to require education for everyone determined in need. GAIN also requires continuous and active participation in the program for as long as someone remains on welfare. Other states are beginning to implement the JOBS "learnfare" provisions, requiring teenage AFDC parents to enroll in school or have their grants reduced. To the extent that these and other state JOBS initiatives are carefully assessed, the promised expansion of employment programs should provide important new information. 16 Uncertainties about how states will use the JOBS resources, about the actual scale of future programs, and about the relative effectiveness of different approaches suggest caution in predicting the results of the new legislation. The lessons from the early to mid-1980s suggest that these programs can improve the tradeoff within the welfare system between the two goals of reducing poverty and promoting self-sufficiency. As a result of these programs, more people are working and a greater share of their income comes from their own earnings, rather than transfers. However, while this did not come at the price of increasing poverty, it was not accompanied by any notable reduction in it, either. While more intensive or targeted services may increase average impacts and overall income, other research suggests that employment programs alone are unlikely 16 Several multi-site social experiments are under way that should provide reliable information on alternative programs for young mothers: a project testing mandatory employment-related activities and case management in two locations (the U.S. Department of Health and Human Services' Teenage Parent Demonstration); a demonstration of a comprehensive voluntary program providing employment, education, health, and child care services in a multi-generational model (MDRC's New Chance Demonstration); and an evaluation of Ohio's statewide learnfare initiative, Project Learn. The U.S. Department of Health and Human Services has also funded a 10-site social experiment to test the impact of different JOBS approaches.