APPENDIX L. MONITORING THE EFFECTS OF PRE- AND POST-TANF WELFARE REFORM INITIATIVES CONTENTS Introduction and Summary of Findings Overview Summary of Findings Implications for TANF Economic Status of Single-Mother Families Overview Cash Welfare Receipt Rates of Employment Work and Welfare Among Poor Single Mothers Effects of Earnings, Transfers, and Taxes on Single Mothers' Poverty Earnings-Poor Single Mothers after Taxes and Transfers Income Sources Among Poorest Single Mothers Consumption Expenditures Findings from Impact Studies of Welfare Reform Initiatives Methodology Issues in Impact Studies Welfare-to-Work Impacts Family Formation and Structure Impacts Economic Status Impacts Child Well-Being Impacts Cost Benefit Issues Findings from Leaver Studies Monitoring Studies Welfare Leavers Ongoing and Future Research Efforts by Governmental Agencies Efforts by Other Organizations Ethnographic Studies Detailed Tables References INTRODUCTION AND SUMMARY OF FINDINGS Overview This appendix summarizes what has been learned from major welfare reform efforts since the mid-1980s to promote work and responsibility by parents of needy children who receive cash aid. During this period, two major Federal welfare reform laws were enacted, the Family Support Act of 1988 (Public Law 100- 485) and the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (Public Law 104-193). In addition, there were numerous State and local welfare reform initiatives. Welfare programs and pilot projects have tested such reforms as work requirements, rewards and penalties, time limits on benefits, education and training provisions, and personal behavior provisions (for example, rules to promote marriage, school attendance, and immunization, and to penalize additional births). Some projects offered enriched support services. The object of this experimentation was to find out what was successful in achieving the goals of public assistance programs for needy families with children, especially the goal of self-support. Congress and welfare administrators wanted to know what worked and for whom and at what cost. Findings reported here are based primarily on programs and initiatives that preceded the 1996 law, which replaced the program of Aid to Families with Dependent Children (AFDC) and its Job Opportunities and Basic Skills (JOBS) Training Program with block grants for Temporary Assistance for Needy Families (TANF). The pre-1996 State reforms tested many policy changes now found in State TANF Programs. The original (1935) purpose of AFDC was to encourage care of needy children in their homes. Over the years, the goal of aiding single-mother families changed, as the movement of American mothers into the work force changed perceptions of mothers' employability and as frustration mounted over the size and character of AFDC rolls. Added to the AFDC law in 1956 were the goals of strengthening family life and promoting ``family self-support.'' To encourage self-support, Congress later created two work and training programs, the Work Incentive Program in 1967 and JOBS in 1988 (Family Support Act). Initially, Congress permitted States to decide who must participate in the Work Incentive Program, but in 1971 it required States to register mothers unless they were caring for a child under age 6. In 1988, Congress tightened the work rule, exempting from JOBS mothers with a child under age 3 (and permitting States to require mothers to work upon the first birthday of their youngest child). The Reagan, Bush, and Clinton administrations gave many States permission to experiment with welfare by waiving specific AFDC rules. (These waivers allowed States to impose restrictions on the receipt of benefits, increase financial work incentives, and establish ``personal responsibility'' requirements for recipients.) TANF has given States much more freedom to design their own programs. Many States have built their TANF Programs on elements of their waiver programs and some have converted their welfare-to-work waiver programs into their TANF Program. Thus, while the evaluations discussed in this appendix are primarily of programs that preceded the 1996 law, some of the findings reported in these evaluations have application to TANF Programs. The 1996 law also broadened the goals of family welfare. TANF has four objectives: (1) enable needy parents to care for children in their homes; (2) end dependence of needy parents on government benefits by promoting job preparation, work, and marriage; (3) prevent and reduce out-of-wedlock pregnancies; and (4) encourage formation and maintenance of two-parent families. Unlike JOBS, TANF makes no exemptions from work requirements, but it allows States to exempt those with a child under age 1. States may exempt others if they wish, but those exempted must still be counted in the denominator when work participation requirements are calculated. To learn from AFDC/JOBS/TANF Programs and experiments, Congress has mandated and supported national data collection efforts as well as evaluations of specific State or local initiatives. It has even authorized specific demonstrations, including major reform projects in Washington (Sec. 9121 of Public Law 100-203), New York (Sec. 9122 of Public Law 100- 203), Minnesota (Sec. 8014 of Public Law 101-239), and Wisconsin (Sec. 233 of Public Law 103-432). All these projects included research designs. To help assess the effects of changes in welfare policy and programs, this appendix offers three complementary analyses from both national data sources and State or local research efforts. 1. The first is an analysis of national data about female- headed families with children, which provides the overall setting. It examines single mothers' receipt of welfare, their work, and their economic well-being. It describes national trends from 1987 to 1998, a period during which the national AFDC/TANF caseload soared to a historic peak in fiscal year 1994 and then dropped by nearly half by 1999. However, this analysis does not attribute changes over time in welfare receipt, employment, or economic well-being of single mothers in the general population to any cause, including welfare policy. 2. The second analysis is a synthesis of findings from evaluations of some of the State welfare reform initiatives. These evaluations examine the difference that policy changes made on selected outcomes, or their ``impact.'' They do so by comparing outcomes, such as earnings and employment rates, under a set of new policies versus what would have occurred without those new policies. Though impact evaluations are designed to measure the effect a program has on selected outcomes, their results are most valid for the particular place and time in which the evaluation was conducted and the population examined in the study. These evaluations were conducted on programs begun before enactment of the 1996 welfare reform law, but many have implications for current TANF Programs. 3. The third analysis is an examination of studies of former cash welfare recipients (welfare ``leavers''). Like the national data, these studies do not attribute their findings to welfare policy changes as they do not make comparisons with groups that were under prior policies. Nor do they cover the full population of interest. Nonetheless, they are a population of particular interest and ``leaver studies'' constitute the bulk of research available since enactment of TANF. Summary of Findings This appendix focuses on outcome measures, which are indicators of how families with children are faring. We focus on four categories of outcomes: 1. Welfare to work, including employment rates, earnings, welfare benefit amounts, and welfare recipiency rates; 2. Family formation and structure, including marriage rates, birth rates, and abortions; 3. Economic status, primarily measured by income, with supplementary information on consumption of goods and services and measures of material well-being and hardship; and 4. Child well-being, primarily measures of a child's development and behavior. Welfare to work The employment rate of single mothers rose steadily from 57 percent in 1992 to almost 71 percent in 1999 (chart L-5), and their employment rate now exceeds that of married mothers. The rise in work by single mothers is especially dramatic for those with a child under age 3; their employment rate was relatively flat (hovering around 35 percent) from 1988 until 1993, then turned upward and climbed to 56 percent by 1999 (chart L-6). At the same time, fewer mothers, even very poor ones, are receiving cash welfare. In the 1987-93 period, cash welfare was received by about 63 percent of mothers who were poor on the basis of pretransfer cash income; by 1998, the share dropped to 41 percent (chart L-3). The overall trend of increasing work and declining welfare is consistent with policies that increase mandated work or job preparation for welfare mothers. Impact evaluations show the same trend. Evaluations find that mandatory welfare-to-work programs generally increase employment, and often also increase average earnings and decrease cash assistance payments. These findings apply both to programs with a strong employment focus (``work first'') as well as to programs that provide education. Of course, the strong economy over this period also had a positive impact on the employment of single mothers. Family formation and structure The number of single-mother families increased from 8.4 million in 1989 to about 9.9 million in 1993, but remained relatively constant from 1993 to 1998, averaging between 9.8 and 10 million during those years (chart L-2). Some think this population ceased growing because of changed attitudes and policies, but reasons are unclear. As noted above, one of TANF's goals is to promote formation and maintenance of two-parent families; another is to reduce out-of-wedlock pregnancies. In fiscal year 1996, the last full year of AFDC, 60 percent of AFDC children for whom data were available lived with a single parent who had not married the other parent, and 13 percent lived with two parents (25 percent were with a divorced or separated parent, and 2 percent with a widowed parent). Many of the evaluations did not examine family formation issues. The majority of pre-TANF evaluations of welfare initiatives that examine marriage report no impacts for single-parent families. Evaluators did find that the Minnesota Family Investment Program (MFIP) decreased the percentage of marriage breakup and increased the rate of marriage among single or cohabiting parents. One popular initiative aimed at reducing nonmarital pregnancies is adopting a family cap (paying no benefit or a reduced benefit for a new baby born to a mother already on welfare), but evaluations of programs that include family caps show inconclusive impacts. Evaluations of programs targeted on teenage AFDC mothers found no significant impacts on rates of pregnancies or childbearing. Economic status Overall, from 1987 to 1998, income has grown and poverty has decreased among mothers raising children alone (data from the Current Population Survey (CPS) of the U.S. Census Bureau). These trends reflect, in part, increased work and less reliance on welfare among these mothers. However, among the poorest mothers (the bottom 20 percent illustrated in chart L-9), income has fallen because these mothers have lost more in cash welfare and food stamps than they have gained in earnings and the earned income credit (EIC). Many impact evaluations find that even welfare-to-work programs that succeed in moving recipients to jobs often do not raise family income. For instance, the National Evaluation of Welfare-to-Work Strategies (NEWWS), which covered 11 JOBS Programs in 7 sites, found no increases in combined income from earnings, cash assistance, and food stamps in the second year of the programs. In these programs, increases in earnings offset, but did not exceed, reductions in cash benefits and food stamps. However, the composition of income changed. A bigger share came from earnings, a smaller share from welfare. However, programs that combine mandated participation in employment-focused activities with generous earnings disregards (which permit working recipients to keep more of their welfare benefits) have been effective in raising earnings and employment. Two of these programs (MFIP and Connecticut's Jobs First) with especially generous treatment of earnings and relatively high welfare payment levels have also raised total income of participants, and the MFIP Program reduced the incidence of poverty. However, their income gains have come at the cost of increasing welfare payments and prolonging the duration of welfare payments. Continued welfare receipt may convey continued health insurance through Medicaid for some participants. Because these programs allowed participants to keep more of their benefits while they worked, they could be said to increase ``dependence'' on welfare. However, because they decreased the share of participants who relied on welfare alone, they also could be said to decrease welfare dependence. Studies of former welfare recipients conducted by the States (``leaver'' studies) address the question of how those who have left welfare are faring. Most families who left TANF or AFDC waiver programs between 1995 and 1998 did so because of employment. In the quarter after exit, administrative data indicate that employment rates ranged from 50 to 64 percent. Among welfare leavers who worked, survey data indicate that average hourly wages ranged from $5.50 to $8.16. Average wages in these States are above the Federal minimum wage and are above the welfare guarantee (cash and food stamps) for the State. However, income from these wages alone would leave a family of three below the poverty level in most States. Within 1 year or at the time of the leaver study, from 13 to 36 percent of leavers had returned to welfare. This count excludes ``churners,'' persons who returned to welfare within 1 or 2 months of exit. Measuring the economic well-being of a family raises some issues. The CPS uses reports of annual family income to measure income and poverty. However, analysis of income and spending data from the U.S. Bureau of Labor Statistics Consumer Expenditure Survey (CEX) shows that some of the poorest single mothers actually buy more in goods and services than would be supported by their income. This finding suggests that they have access to funds from sources other than reported income (e.g., borrowing, withdrawing savings, other family members, boyfriends, or informal support networks). Impact evaluations also encounter problems in measuring economic well-being. The most common measure is the combination of a program participant's earnings, cash welfare, and food stamps. This approach ignores the income of other household members (e.g., husbands, boyfriends, roommates) who might pool their economic resources with the participant, as well as other sources of income such as Social Security or Supplemental Security Income (SSI). Leaver studies have begun to report measures that examine whether welfare leavers have experienced any types of ``hardship.'' Measures of hardship have included receipt of other types of assistance to complement earnings, as well as measures that examine difficulties experienced since exiting cash welfare. Child well-being In recent years, the proportion of AFDC/TANF adults who work has more than doubled, from 11 percent in fiscal year 1996 to 23 percent in fiscal year 1998, and there is growing interest in how increased work by these parents affects their children. The NEWWS evaluation found small and conflicting impacts, or no impact at all, on various measures of child well-being for respondents with only school-aged children. For instance, 4 of the 11 programs were reported to increase the percentage of children who attended a special class for behavioral or emotional problems; 2 of the programs were reported to decrease this percentage. The MFIP Program that increased family income was found to decrease the incidence of problem behaviors; it also improved school performance. The New Hope Program in Milwaukee, which failed to increase family income, was found to improve the behavior of boys, but to increase aggression among girls. Implications for TANF Findings presented in this appendix indicate that ending dependence on government benefits through job preparation, work, and marriage--the second objective of TANF--may be an elusive goal. Available evaluations generally show that welfare initiatives have little or no impact on marriage. And although welfare-to-work programs generally increase earnings and employment, they do not necessarily end dependence and raise income. Programs that increase incomes instead have been found to reduce dependence on welfare; they continue at least partial welfare payments when recipients go to work. Evaluations indicate that, at least in the short run, many TANF recipients will be unable to escape poverty through work unless taxpayers provide an earnings supplement of some kind (see table 7-12 in section 7). TANF's lifetime time limit on eligibility for federally funded assistance conflicts with enhanced earnings disregards. The time limit may induce some persons to leave welfare quickly and ``bank'' welfare months for later use in time of need (although research has not found this effect), but generous earnings disregards are an incentive to remain on welfare while working. The disregards permit recipients to increase income by combining work with a reduced welfare grant. Generous earnings disregards also are an incentive to join the rolls, and they may reward persons who would have worked anyway, even if all earnings were used to reduce the grant. Eventually, when a family reaches the Federal time limit, the State can no longer pay an earnings supplement unless it uses its own funds to pay the cost (or grants the family a hardship time limit extension, within a Federal caseload cap set in the law). Under TANF, the $16.5 billion of Federal funding received by States for aid to needy families is fixed each year through fiscal year 2002. Federal welfare funding no longer is a function of the caseload size, the benefit levels paid, the generosity or stringency of program rules, or the State's relative per capita income. Because the national TANF caseload now is only about half the size of the record-large AFDC caseload on which the fixed block grant is based, States generally have available much larger Federal funding per welfare family than before TANF. States may use these funds for a wide array of aid and services, not restricted to cash recipients, provided they are related to one of TANF's goals. Rules about State funding have changed under TANF and also give the States more discretion. No longer do States have to pay a specified share of benefit expenditures, depending on per capita income, and a flat share of administrative costs. Instead, they must meet a ``maintenance-of-effort'' requirement of 75 or 80 percent of their spending in a recent baseline year ($10.4 billion yearly, or $11.1 billion if they fail work participation rates). These amounts are below fiscal year 1994 State AFDC/JOBS spending levels by 25 percent and 20 percent, respectively. Many welfare reform initiatives, such as the operation of a welfare-to-work program, cost Federal and State governments money. The costs of welfare reform initiatives can be recouped if over the long run welfare payments are reduced. Employment- focused (``work first'') programs, on average, have lower costs than those that promote education. Thus, it is more likely that costs of a ``work first'' program can be recouped than those of an education-focused program. Moreover, the benefits from a ``work first'' program tend to be immediate, while those from an education-focused program take time to emerge. Additionally, earnings supplements that encourage work and raise incomes also cost money; they prolong the duration of welfare. Fixed funding is often viewed as an incentive for States to operate less expensive programs. However, the TANF Block Grant is fixed at a level based on historically high caseloads. Should caseloads remain down, States would continue to have the resources to engage in more costly welfare initiatives, as many of them appear to be doing now. ECONOMIC STATUS OF SINGLE-MOTHER FAMILIES Overview This section examines trends in welfare, work, and economic well-being among families headed by single mothers, the group that is the main focus of TANF Programs. The analyses use national household survey data, specifically, the U.S. Census Bureau's March CPS and the Bureau of Labor Statistics' CEX. CPS data were examined from March 1988 (income year 1987) to March 1999 (income year 1998). CEX data were examined from 1994 through 1997. (The most recent year for which public use files were available in early 2000, when this analysis was completed, was 1997.) The analyses seek to describe the changing circumstances of single mothers over the period, but do not attempt to isolate the effects of particular policy changes or to ascribe causation. These analyses are presented to provide background and context for the subsequent discussions of research focused specifically on welfare reform initiatives. Given the precision of estimates obtained from the CPS and CEX, this section focuses primarily on overall trends, rather than specific year- to-year changes. The number of single mothers grew by 17 percent over the 4- year period from 1989 to 1993, but has since leveled off and remained at around 10 million (chart L-2). Dramatic changes in work, welfare, and poverty among this population have occurred in recent years, especially since 1992-93, as detailed in the following charts. Highlights include: --The percentage of single mothers who worked at some time during the year rose from 67 percent in 1992 to 80 percent in 1998 (chart L-1). By March 1998, the employment rate for single mothers with children under age 18 surpassed that of comparable married mothers (chart L-5). Employment gains have been greatest for single mothers with children under age 3 (chart L-6). --The percentage of single mothers who received cash welfare, based on CPS data, shrank from 35 percent in 1993 to 19 percent in 1998 (chart L-1). --The percentage of poor single mothers (on the basis of pretransfer income) who reported receiving cash welfare declined from 63 percent in 1993 to 41 percent in 1998 (chart L-3). Declines occurred even among those with very low pretransfer income (e.g., below 25 percent of poverty) (chart L-4). --The percentage of single mothers who were poor based on money income, after cash transfers (the official poverty measure), declined from 45 percent in 1993 to 37 percent in 1998 (chart L-1 and chart L-8). If in- kind food assistance and the EIC (net of taxes) were counted as income, the poverty rate would have dropped from 41 percent in 1993 to below 30 percent in 1998 (chart L-8). --1997 marked a transition year in which, for the first time over the period examined, the share of poor single mothers who worked during the year exceeded those who received welfare (chart L-7). --Average total income of single mothers in the bottom quintile declined from 1994 to 1998, despite a gain of $1,005 in net earnings and EIC transfers, because of a greater decline ($1,788) in cash welfare and food stamps (chart L-10). However, in the second lowest quintile, gains in earnings and the EIC by single mothers offset declines in cash welfare and food stamps, resulting in higher total income in 1998 than at any time during the preceding 11 years for this subset of the population (chart L-11). --Between 1994 and 1997, consumption expenditures for single- mother families grew at a robust rate: 18 percent over the period, much higher than the increase in prices over the same period (8 percent). Some of this increase is attributable to work-related expenses (transportation, child care, payroll taxes), as welfare receipt decreased and work increased (table L-1). Consumption expenditures increased even for the poorest single mothers (chart L-12). CHART L-1. WELFARE, WORK AND POVERTY STATUS AMONG SINGLE MOTHERS, 1987- 98 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Cash Welfare Receipt CPS data show an increase in cash welfare receipt (AFDC, TANF, or other assistance) among single mothers during the late 1980s and early 1990s and a decrease in the mid-to-late 1990s that corresponds to the caseload rise and fall documented by administrative data. Chart L-2 shows that the total number of single mothers increased from 8.4 million in 1989, to about 9.9 million in 1993, an increase of 1.5 million, or 17 percent. Since 1993, the number of single mothers has remained fairly stable, between 9.8 and 10 million, but the number of single mothers receiving cash welfare has fallen each year. CHART L-2. SINGLE MOTHERS: POVERTY AND CASH WELFARE RECEIPT, 1987-98 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. The number of single mothers in families receiving cash welfare increased from 2.5 million in 1989, to 3.4 million in 1993, an increase of 900,000, or 36 percent over the 4-year period. Since 1993, the number of single mothers reporting cash welfare has fallen to 1.9 million (a 44 percent decline). Over the same period, the number of poor single mothers who reported receiving no cash welfare increased by 532,000, from 1.721 million in 1993 to 2.253 million in 1998 (the middle shaded area shown in chart L-2). Chart L-3 shows that cash welfare recipiency rates among single mothers overall, and among poor single mothers based on their pretransfer income (cash income excluding cash welfare), remained fairly steady during the 1987-93 period, but have fallen considerably since. Among single mothers who were poor based on their pretransfer cash income, the share who received cash welfare generally hovered around 63 percent over the 1987- 93 period. As the chart shows, the likelihood of cash welfare receipt has decreased since 1993 among this population. In 1993, the cash welfare recipiency rate among single mothers with pretransfer income below poverty was 63 percent; by 1998, it had fallen to 41 percent. CHART L-3. SINGLE MOTHERS: CASH WELFARE RECIPIENCY RATES, 1987-98 * Pretransfer income is cash income other than cash welfare payments. Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Chart L-4 shows cash welfare recipiency rates based on families' pretransfer (i.e., precash welfare) income relative to poverty. The top line of the chart depicts the share of single mothers without other income who reported receiving cash welfare. The line shows that nearly 90 percent of single mothers with no pretransfer income reported receiving cash assistance in 1987-90. However, since 1990, the reported rate of cash welfare recipiency for this group has continuously declined, to 77 percent in 1996, and to 61 percent by 1998. Similarly, for families with very low pretransfer income (below 25 percent of poverty), and for families with pretransfer incomes between 25 and 50 percent of poverty, cash welfare recipiency also shows dramatic declines: for the former group from 72 percent in 1996 to 56 percent in 1998, and for the latter group from 60 percent in 1995 to 45 percent in 1998. CHART L-4. CASH WELFARE RECIPIENCY RATES AMONG SINGLE-MOTHER FAMILIES BY PRETRANSFER INCOME POVERTY STATUS, 1987-98 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Likewise, food stamp recipiency rates among low-income households have also fallen considerably since 1994. In 1994, 71 percent of single-mother families with household income below 130 percent of poverty (the Food Stamp Program's gross income qualifying limit) reported receiving food stamp benefits; by 1998, the share fell to 57 percent. Among those with household incomes below 50 percent of the low-household income threshold, in 1994, 80 percent reported food stamp receipt; in 1998, 70 percent reported food stamp receipt. Rates of Employment While welfare receipt has declined, dramatic gains in single mothers' employment have occurred since 1994. Chart L-5 shows that for most of the period shown, the employment rate for single mothers with children under the age of 18 was below that of comparable married mothers. However, by March 1998, the employment rate of single mothers surpassed that of married mothers. The employment rate among single mothers rose steadily from a recent low of about 57 percent in March 1992 and 1993, to about 71 percent in March 1999. Most dramatic has been the increase in employment among single mothers who have a child under the age of 3; their employment rate increased from a recent low of 35.1 percent in March 1993 (18.1 percentage points below the rate for married mothers) to a high of 55.8 percent in March 1999 (just 1.2 percentage points below their married counterparts) (chart L- 6). Single mothers with a youngest child age 3-5 also experienced marked employment gains over the mid-to-late 1990s. Their employment rate grew from a recent low of 54.1 percent in March 1992, to 69.8 percent by March 1999, a 15.7 percentage point increase, surpassing that of their married counterparts by 6.7 percentage points (not shown). Single mothers whose youngest child was of school age (age 6-17) had employment rates about equal to those of their married counterparts over the 1988-99 period (not shown). CHART L-5. EMPLOYMENT RATES OF MARRIED AND SINGLE MOTHERS WITH CHILDREN UNDER AGE 18, MARCH 1988-MARCH 1999 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. The healthy economy, combined with a transformed welfare system, improvements in the EIC, and increases in the minimum wage, are among the factors thought to have encouraged work among single mothers in recent years. TANF, and the AFDC waivers that preceded it, transformed cash assistance from a needs-based entitlement to a program of temporary assistance, encouraging work and personal responsibility. Imposition of work requirements, time limits, and sanctions, and in some States, more generous earnings disregards, all serve to encourage work, either in lieu of welfare, or for a temporary period, in conjunction with welfare. The EIC, which is conditioned on earnings, is thought to encourage work among most groups, especially single parents who were not working, or who were marginally attached to the labor market. Increases in the EIC, passed by Congress in 1993 and phased in between 1994 and 1996, have increased the financial incentive for many single mothers to work. Other factors, such as increased funding for child care subsidies, may also have contributed to making work possible for more single mothers. CHART L-6. EMPLOYMENT RATES OF MARRIED AND SINGLE MOTHERS WITH CHILDREN UNDER AGE 3, MARCH 1988-MARCH 1999 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Work and Welfare Among Poor Single Mothers As shown in the previous section, poor single mothers are less likely to be receiving cash welfare than in previous years. Likewise, like all single mothers, poor single mothers are now more likely to be working. Changes in poor mothers' participation in work and welfare first became evident in the early-to-mid 1990s, with rates of welfare receipt declining after 1993, and rates of employment increasing after 1992 (see chart L-7, top 2 lines). A crossover point was reached between 1996 and 1997, when the chances that a poor single mother would be working exceeded the chances that she would be receiving welfare. Chart L-7 shows that the share of poor single mothers who received cash welfare at any time during the year fell from about 60 percent in the 1987-93 period, to just over 40 percent in 1998. The rate of decline in welfare receipt among poor single mothers was greatest between 1996 and 1997, a period coinciding with the passage and implementation of national welfare reform legislation. Similarly, the share of poor single mothers who were working at any time during the year increased from just above 40 percent in 1992, to about 53 percent in 1998, with the greatest increases occurring in 1996 and 1997. The share of poor single mothers who relied on cash welfare without working dropped from a peak of 45 percent in 1991, to about 22 percent in 1998, while the share who worked without relying on cash welfare has increased from a recent low of about 23 percent in 1993, to just above 34 percent in 1998. Most of the increase in work without welfare occurred in 1996 and 1997. The share of poor single mothers who combined work and welfare over the year has remained relatively constant, between 17 and 19 percent, except for 1 year (1989, when the share was about 15 percent). CHART L-7. POOR SINGLE MOTHERS: WORK AND WELFARE STATUS DURING THE YEAR, 1987-98 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. The share of poor single mothers who reported that they neither worked, nor received cash welfare during the year (the dashed line in chart L-7), has increased each year since 1993 (15.2 percent), reaching nearly 25 percent in 1998. This surprising combination may reflect a mix of circumstances, including income or support from other sources such as family members, support from unrelated household members (which is not included in the official poverty measure), and other means of support from outside the household not counted on the CPS. It may also reflect income reporting problems on the CPS, especially with regard to welfare income. \1\ Finally, welfare sanction and diversion policies may have contributed to the increased number of poor mothers neither working nor receiving welfare. --------------------------------------------------------------------------- \1\ A comparison of AFDC/TANF administrative statistics and CPS- estimated caseload counts suggests that the CPS undercounts actual cases, and that the CPS undercount has worsened in recent years. From 1987 to 1991, the CPS accounted for roughly 80 percent of the AFDC administrative caseload count, but by 1998 the CPS was capturing only about 64 percent. Worsened reporting of cash welfare on the CPS makes it difficult to gauge how much of the drop in welfare receipt among single mothers represents eligible families who do not receive assistance, rather than families who do not report actual welfare aid on the CPS. To at least some extent, the declining welfare recipiency rates discussed in this section are likely due to increased underreporting of cash welfare on the CPS. See Bavier (2000) for a detailed discussion of cash welfare underreporting on the CPS and other surveys. --------------------------------------------------------------------------- Effects of Earnings, Transfers, and Taxes on Single Mothers' Poverty As shown in chart L-1, single mothers' poverty status has improved since 1993. Changes in the economy and changes in welfare policy and other programs, such as the EIC, have both direct and indirect effects on poverty. However, the official U.S. poverty measure counts only family cash income (excluding capital gains and lump-sum or one-time payments) against a family's poverty threshold, which varies by family size and composition, to determine whether a family is counted as poor. The definition does not include the value of in-kind benefits, such as food stamps, school lunches, or public housing subsidies, nor does it include the effects of taxes or tax credits such as the EIC. Inclusion of in-kind benefits and the EIC provides a more comprehensive income definition than the official definition. Additionally, other unrelated household members may contribute to the family's economic well-being, but determining the extent to which resources are shared among unrelated household members is often difficult. Chart L-8 shows the effects of income from these other sources on poverty among all single mothers. Components of family income are sequentially added and measured against families' poverty thresholds, as one moves from the top line of the chart to subsequent lines below: CHART L-8. EFFECTS OF EARNINGS, TRANSFERS, AND TAXES ON FAMILY POVERTY AND HOUSEHOLD LOW-INCOME STATUS ON SINGLE MOTHERS, 1987-98 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. --Line 1: The top line shows the percent of single mothers who would be counted as poor if only family earnings were counted against the poverty line. --Line 2: The second line down includes other sources of cash income, in addition to earnings, that were already counted above. However, this line does not include cash welfare. --Line 3: The third line down adds cash welfare to the other sources already mentioned, and with those sources, represents the income definition used in the official poverty measure. --Line 4: The fourth line down shows the value of in-kind food assistance (i.e., food stamps, free and reduced price school lunches, Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) payments) when added to cash income and compared to the family poverty threshold. --Line 5: The fifth line down shows the effect of adding the value of the EIC, less Federal and State income taxes and payroll taxes, to line 4. --Line 6: The bottom (dashed) line shows the effects of counting all income in the household in which the single mother lives, not just that of her own family, and comparing it to an unofficial ``household low- income threshold.'' The household low-income threshold used here applies family poverty income thresholds, which are based on family size and composition, to households, based on household size and composition. It must be noted that official poverty measurement is based on a family concept, which assumes that family members share income and economies of scale that result from shared living arrangements. It is generally agreed among researchers that assumptions regarding income sharing and shared economies of scale among related family members, who have ties based on blood, marriage, and adoption, do not apply to the same extent among unrelated household members. Consequently, these estimates of household low-income status likely overstate the effect of household income on reducing poverty among families headed by single mothers. In viewing chart L-8, note that the trend in earnings is the principal factor affecting the declining trend in poverty, whereas the other income sources, with the exception of the EIC, affect the level of poverty, more than its trend over time. Evidence of this effect is that most lines in the chart, with the exception of the EIC, roughly run parallel to the ones above. Effect of earnings and other nonwelfare cash income on poverty Chart L-8 shows that between 1993 and 1998, single mothers' poverty, based on family earnings alone, fell from 56.2 percent to 47.9 percent (line 1). Adding other cash income, except cash welfare, to family earnings (line 2), reduces poverty in 1993 from 56.2 percent to 47.4 percent, and in 1998, from 47.9 percent to 38.8 percent. Effect of cash welfare on poverty When added to other income, cash welfare benefits have only a small impact on the poverty rate, as these benefits generally are not sufficient, even when combined with other cash income, to lift families above the Federal poverty threshold. As shown by table 7-13 (breakeven points) in section 7, in all but eight States, Temporary Assistance for Needy Families (TANF) families that go to work lose eligibility for benefits before earnings reach the poverty line, usually long before. Consequently, cash welfare benefits have little impact on the poverty rate. The addition of cash welfare (line 3, representing the official income definition for measuring poverty) reduces poverty only slightly, from 47.4 percent (line 2) to 45.2 percent in 1993, and from 38.8 percent to 37.3 percent in 1998. Nonetheless, cash welfare benefits can have a significant impact on the level of poor families' incomes, affecting the degree to which their incomes fall below the poverty income standard. This impact is not captured by changes in the poverty rate. Effect of in-kind food assistance on poverty The fourth line from the top in chart L-8 shows the effect on the poverty rate of single mothers by counting government food assistance, in the form of food stamp benefits and school lunch benefits and WIC payments. The line shows that food assistance reduces the poverty rate of single mothers from about 2-3 percentage points over the period. The antipoverty effectiveness of food assistance seems to have lessened somewhat in recent years. In 1995, food assistance reduced the poverty rate from 40.2 percent (its official measure) to 36.9 percent, a 3.3 percentage point (8.1 percent) reduction in poverty. In 1998, food assistance reduced the poverty rate from its official rate of 37.3 percent, to 35.2 percent, a 2.1 percentage point (5.6 percent) reduction. Effect of EIC and taxes on poverty As noted above, the net effect of the EIC (after counting the effect of reductions in income from Federal and State income taxes and FICA taxes) (line 5), when added to total family cash income and food assistance (line 4), causes a divergence in trend from the lines above. This is especially notable after 1993. A major expansion of the EIC, passed by Congress in 1993 and phased in between 1994 and 1996, increased the amount of the EIC work bonus families might receive. The antipoverty effectiveness of the EIC was approximately three times greater in 1998 than in 1993. In 1993, the EIC reduced the poverty rate (counting food assistance) among single mothers from 42.7 percent (line 4), to 40.7 percent (line 5), a 2.0 percentage point (4.6 percent) reduction. In 1998, the EIC reduced poverty from 35.2 percent to 29.7 percent, a 5.5 percentage point (15.7 percent) reduction.\2\ --------------------------------------------------------------------------- \2\ Note that the value of the EIC on the CPS is based on U.S. Census Bureau imputations, rather than actual reported tax credits. Also, the EIC is different than most sources of income, as most families receive the EIC as a lump-sum refund. --------------------------------------------------------------------------- As receipt of the EIC is conditioned on earnings, the growing impact of the EIC in part reflects the rise in work rates among single mothers. Among those who are working and poor (before counting the EIC), the EIC helps lift the income of some above the poverty line. Although the EIC expansion provided additional income to low-income families who were already working, it may also have helped induce increased employment among family heads with low to moderate earnings potential, and thus contributed to the decline in poverty based on earned income only that has occurred since 1993 (shown as the top line in the chart). Note too, that to the extent that changes in cash welfare programs in recent years have encouraged work, these changes may have had an indirect effect on poverty by increasing earnings and, through earnings, making the EIC available to a greater number of families. Effect of all household income on poverty The household low-income line (bottom line) shows that if all household members' income were shared equally among household members, the poverty rate among single mothers would drop by at most 3-4 percentage points over the 1987-98 period. Adding other members' household income, and counting them as though they were family members who shared income equally, reduced the post-in-kind transfer, posttax, poverty rate in 1993 from 40.7 percent to 36.8 percent; in 1998 the post-in- kind transfer, posttax, poverty rate would have dropped from 29.7 percent to 26.2 percent. Again, this is most likely an overstatement of the possible effect that shared household living arrangements might have on single mothers' poverty status, because of uncertainty about the extent to which such income is actually shared. Earnings-Poor Single Mothers After Taxes and Transfers This section focuses specifically on single mothers with family incomes below poverty, based on family earnings alone, to gauge the effects of other income, transfers and taxes over time. This group, shown in the top line in chart L-8 for the 1987-98 period, accounted for about half of all single mothers. Chart L-9 shows their distribution, relative to the poverty threshold, after cash income from all sources, in-kind food assistance, and taxes (including the EIC) is taken into account. Unlike chart L-8 above, which simply measured whether income was below poverty, chart L-9 examines the degree of poverty after taxes and transfers, for families defined as poor based on family earnings alone. The chart includes families with no earnings, as well as those who have earnings but whose earnings fall below poverty. The chart shows that while the majority of ``earnings- poor'' single mothers have seen improvements in income relative to poverty since 1991, the improvement occurred mostly in 1994 and 1995 (although it continued after a slight decline in 1996). Since 1996, the bottom 40 percent of these single mothers have seen no improvement in their income relative to poverty, and the bottom 20 percent have actually seen declines in their income status relative to poverty. The chart shows, for example, that over the 1987-98 period, the top 20 percent of single mothers who were poor based on earnings alone had income from other sources that helped bring their families' incomes above the poverty line. Over the period, these families became somewhat more economically secure: in 1987, the top 20 percent of earnings-poor mothers had a net in-kind aftertax income that was 5 percent or more above the poverty line. By 1998, the top 20 percent of earnings-poor single mothers had net income that was 25 percent or more above poverty. The chart shows strong net income gains relative to poverty among earnings-poor single mothers over the 1993-95 period, followed by a slight decline in 1996, for all but the bottom 20 percent of mothers. Since 1996, mothers in the top half of the distribution show income gains relative to poverty. However, those in the bottom 40 percent show no gain in income security since 1996, and those in the bottom 20 percent show a decline in their level of income security. The bottom 20 percent of earnings-poor single mothers show a comparatively large decline in net income relative to poverty in 1997 (from 52 percent of poverty in 1996, to 46 percent of poverty in 1997 and 1998). CHART L-9. POSTTAX POST-IN-KIND TRANSFER INCOME AS A PERCENT OF POVERTY AMONG SINGLE MOTHERS CONSIDERED POOR BASED ON FAMILY EARNINGS ALONE, 1987-98 Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Income Sources Among Poorest Single Mothers \3\ --------------------------------------------------------------------------- \3\ In addition to the analysis which follows, see: Primus, et al. (1999); Bavier (1999); and Haskins (in press) for other research discussing recent declines in income among the poorest families. --------------------------------------------------------------------------- The composition and level of income among the poorest single-mother families has changed markedly in recent years, reflecting increased earnings supplemented by increased earned income credits (EIC) and reductions in cash welfare and food stamps. For single mothers in the bottom fifth (bottom quintile), increased earnings and EIC have not been sufficient in recent years to offset losses in cash welfare and food stamps, resulting in reduced income since 1996. Families in the bottom 20-40 percent (second quintile) also received less cash welfare and food stamps in recent years, but in 1998, increased earnings and EIC were sufficient to offset these losses. Charts L-10 and L-11 examine sources of income among the bottom quintile (bottom 20 percent) and the second lowest quintile (bottom 20-40 percent) of single-mother families, respectively, based on their pretax cash income relative to poverty. The charts show the average annual income, in 1998 dollars, from the following sources: cash public assistance (Aid to Families with Dependent Children (AFDC), TANF, and general assistance); Supplemental Security Income (SSI); food stamps (market value); child support and alimony; other cash income other than earnings; net earnings (earnings net of the employee share of FICA payroll taxes and any Federal or State income taxes); and the EIC. The employee share of FICA payroll taxes, and any Federal or State income tax payments are also shown as negative values. Note that these estimates are based on year-to-year income comparisons of cross-sectional survey data, rather than a comparison of incomes for the same families over time. CHART L-10. BOTTOM QUINTILE OF SINGLE-MOTHER FAMILIES: AVERAGE ANNUAL INCOME BY SOURCE, 1987-98 (IN 1998 DOLLARS) Note._Quintiles based on ranking of ratios of family cash, pretax income, relative to poverty. Taxes include Federal and State income taxes and FICA taxes. Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Chart L-10 shows an upsurge in annual average income among single mothers in the bottom quintile, from 1993 to 1994. Average total income increased from $6,119 to $6,997; an increase of $878, or 14 percent. Components of the increase were: cash public assistance, $97 (4 percent); food stamps, $257 (12 percent); SSI, $84 (56 percent); net earnings, $316 (50 percent); and EIC, $152 (131 percent). From 1994 through 1996, average total income among the bottom quintile of single-mother families remained essentially unchanged, drifting down slightly perhaps, despite dramatic changes in its composition. In 1995 and 1996, earnings of single mothers in the bottom quintile continued to increase, as did EIC, while cash assistance (AFDC and general assistance) and food stamps fell. The growing importance of the EIC as an earnings supplement over the 1994-96 period can be illustrated by examining the average EIC as a share of average earnings. From 1993 to 1996, the EIC ``work bonus'' had doubled, from 18 percent to 37 percent of earnings. CHART L-11. SECOND QUINTILE OF SINGLE-MOTHER FAMILIES: AVERAGE ANNUAL INCOME BY SOURCE, 1987-98 (IN 1998 DOLLARS) Note._Quintiles based on ranking of ratios of family cash, pretax income, relative to poverty. Taxes include Federal and State income taxes and FICA taxes. Source: Prepared by the Congressional Research Service. Based on analysis of U.S. Census Bureau March 1988-99 Current Population Survey data. Chart L-10 further shows that average cash welfare and food stamp benefits reported by single mothers in the bottom quintile continued to decline in 1997 and 1998, and that earnings combined with the EIC did not rise enough to offset this loss. As a result, average total family income for this population was lower in 1997-98 than in 1994-96; however, it was higher than in all years preceding 1994, with the exception of 1987. Average annual cash welfare assistance received by single mothers in the bottom quintile in 1998 was 43 percent below what this group had received, on average, in the most recent high year of 1994 ($1,410 versus $2,486). Food stamps for this group in 1998 were 29 percent below their 1994 value ($1,710 versus $2,422). In contrast, net earnings were 67 percent higher in 1998 than in 1994 ($1,612 versus $966), and nearly 2\1/2\ times higher than in 1993 ($1,612 versus $650). Nonetheless, while combined cash assistance and food stamps fell by $1,788 from 1994 to 1998, net earnings combined with EIC grew by $1,005, and offset only 56 percent of the loss in cash welfare and food stamps over the period. Chart L-11 is similar to chart L-10, but shows average income by source for the second quintile of single-mother families, ranked by their income relative to poverty. The chart shows comparatively large gains in average total income from 1993 to 1995, due largely to increased earnings and EIC. Over this period, average total income increased from $11,396 to $14,433, a gain of nearly 27 percent. With the exception of 1996, average earnings for single mothers in the second quintile continued to grow; however, earnings and the EIC were insufficient to offset declines in cash assistance and food stamps in 1996 and 1997. From 1995 to 1997, combined earnings and EIC gains ($792) offset only 63 percent of the loss in combined cash assistance and food stamp benefits ($1,253) over the period. By 1998, average total income among single mothers in the second quintile reached a new high. In 1998, earnings in combination with the EIC were more than offsetting the loss in combined cash assistance and food stamps that occurred over the 1995-98 period. The gain in average net earnings, in combination with EIC ($2,343), more than offset the $1,894 loss in combined cash assistance and food stamps. By 1998, average net earnings accounted for over half of these families' incomes ($7,329 in earnings out of a total net income of $14,625) and cash assistance ($1,177) accounted for just 8 percent. In contrast, in 1987, earnings accounted for about 28 percent of this group's income ($3,363 in earnings out of a total net income of $12,013) and cash assistance ($4,157) comprised about 35 percent. In 1998, average total income for families in the second quintile ($14,625) was nearly 22 percent above that in 1987 ($12,013). Consumption Expenditures Though annual income information is the basis of most official measures of economic well-being (e.g., poverty statistics), it suffers some drawbacks. Income during any one period imperfectly measures the value of goods and services a family can consume, since families may liquidate savings, borrow money, or access some other sources of funds. This section presents information on consumer expenditures from the U.S. Bureau of Labor Statistics Consumer Expenditure Survey (CEX), as a complement to the preceding analysis of income data from the Current Population Survey (CPS). Consumption data are derived from the CEX, which has a relatively small sample size of 5,000 households, compared to the CPS which has a sample of 47,000 households. The small CEX sample size, relative to the CPS, means that data from the CPS are likely to be more reliable, and large year-to-year differences are required for changes estimated from the CEX to meet standard tests of statistical significance. Differences noted in the text met tests of statistical significance, using ``replicate sample'' techniques (Mosteller & Tukey, 1977). Average consumption for single mothers Table L-1 summarizes information on consumption expenditures, welfare receipt, and work for single-mother families from 1994 to 1997. The CPS data described above showed that poverty continued to decline for these families over these years. The CEX data show that consumer expenditures grew for this group over the period, by 18 percent. This is well above inflation during the 1994-97 period (8 percent). The CEX information, like the CPS data, shows a sharp decline in welfare receipt and an increase in work among single mothers over this period. TABLE L-1.--CONSUMPTION EXPENDITURES, WELFARE RECEIPT, AND WORK EXPERIENCE FOR SINGLE-MOTHER FAMILIES, 1994-97 ---------------------------------------------------------------------------------------------------------------- Year Percent Category ---------------------------------------- change, 1994 1995 1996 1997 1994-97 ---------------------------------------------------------------------------------------------------------------- Average consumption expenditures............................. $19,633 $19,830 $21,405 $23,245 18.4 Expenditures on ``work-related'' categories................... 4,136 4,361 5,050 5,178 25.2 Other consumption expenditures................................ 15,497 15,469 16,355 18,068 16.6 Welfare receipt (percent)..................................... 33.9 32.8 23.2 20.3 -40.2 Average number of workers in family........................... 0.93 0.92 1.00 1.02 9.8 ---------------------------------------------------------------------------------------------------------------- Note.--Data are from quarterly interview surveys; quarterly expenditures have been annualized for purposes of display; year noted represents year of the interview; actual expenditures may have occurred in the prior year; dollars are not inflation-adjusted. Source: Congressional Research Service tabulations of data from the U.S. Department of Labor, Bureau of Labor Statistics Consumer Expenditure Survey. The table shows overall consumption expenditures for single-mother families but also divides consumption expenditures into: (1) a category that explicitly includes spending on items commonly associated with work (transportation, child care, and retirement contributions including the employee share of Social Security payroll taxes); and (2) all other expenditures (including expenditures to meet basic needs, such as food and shelter). As shown in the table, some of the overall increase in consumption expenditures for single mothers is attributable to work expenses. The work expenses category showed a 25 percent increase over the 1994-97 period, while other expenses grew by 17 percent. The increase in work expenses is consistent with other trends shown in the table, such as the decline in welfare receipt and increase in work. Consumption expenditures for the poorest families Analyses of national income surveys (for example, the CPS) have shown that income growth among the poorest single mothers stagnated after 1994. Information from the CEX also shows that income for the poorest one-fifth of single mothers has not grown since 1994 (the estimates actually show a decline in income, but that decline is too small to meet standard criteria of statistical significance). However, even among the poorest single-mother families, consumption expenditures have increased. Chart L-12 shows aftertax income, consumption expenditures classified as work expenses, and other consumption expenditures for the bottom quintile of single-mother families during 1994- 97. The chart shows that income remained relatively stable, while consumption expenditures--both those commonly thought of as work expenses and other consumption expenditures--increased. CHART L-12. AFTERTAX INCOME, WORK EXPENSES, AND OTHER CONSUMPTION EXPENDITURES FOR THE POOREST FIFTH (INCOME QUINTILE) OF FEMALE-HEADED FAMILIES; 1994-97 Note._Data are from quarterly interview surveys; quarterly expenditures have been annualized for purposes of display; year noted represents year of the interview, actual expenditures may have occurred in the prior year. Information is for ``complete income reporters'' only. Dollars are not inflation adjusted. Source: Congressional Research Service tabulations of data from the U.S. Department of Labor, Bureau of Labor Statistics Consumer Expenditure Survey. The chart also shows that, for the low-income population, reported consumption expenditures typically exceed reported incomes. Over the 1994-97 period, this gap appeared to grow. There are a number of possible explanations: 1. Some who have low incomes are temporarily poor. For those families that are experiencing a temporary dip in income, withdrawing savings can finance consumption expenditures that exceed income. 2. Some low-income people might access credit markets to pay for spending that exceeds income. Poor families increased their credit card debt in the early-to-mid 1990s (Bird et al., 1999) to a greater extent than other families. If this trend continued into the later 1990s, it could account for some of the growing gap in spending and income among the low-income population. 3. Income might be underreported relative to spending, particularly for certain types of income. Edin and Lein (1997) discuss various survival strategies that very low-income, single mothers use to make ends meet: network-based or agency-based support, unreported work and work in the underground economy, and cash contributions from families, friends, boyfriends, or former spouses who live elsewhere. These latter types of ``income'' are unlikely to be reported as income on household surveys. Interpreting the consumption data This analysis of consumption expenditures complements, but does not substitute for, the analysis of income data presented earlier. However, its finding that many of the lowest income families can consume well above their income underscores the difficulties in measuring the economic well-being of single mothers and their families. FINDINGS FROM IMPACT STUDIES OF WELFARE REFORM INITIATIVES Methodology Issues in Impact Studies A large body of research on the effects of welfare reform initiatives is available from impact evaluations. Unlike the national data discussed above, which show trends in outcomes related to welfare reform but do not attribute these trends to any single cause, impact evaluations are designed to measure the difference a program makes on outcomes related to its goals. For example, a welfare-to-work program will be evaluated based on the difference between employment, earnings, and welfare receipt under the program versus the outcomes produced by an alternative set of policies. The focus of an impact evaluation is the incremental difference in outcomes observed under one set of policies versus another. Random assignment experiments There are a number of different techniques available to evaluate the impact of policy changes, but most evaluations conducted after the passage of the Family Support Act in 1988 used random assignment to experimental and control groups. These studies assign potential participants to two or more groups. Individuals assigned to a control group are subject to current policies (no policy change); individuals assigned to the experimental group (or groups) are subject to a different package of policy initiatives, such as time limits on assistance, or sanctions for failing to comply with a mandatory work requirement. Because individuals are randomly assigned to these groups, any differences between the experimental and control groups may be attributed to the policy initiative itself. This difference is therefore the impact of these policy changes. For a policy to have an impact, it must be determined that the impact was not simply a chance occurrence. Differences between experimental and control groups that pass statistical significance tests are reported as policy impacts. Whether a difference is determined to be an impact generally depends on two factors: (1) the size of the difference between the experimental and control groups, and (2) the size of the sample in the evaluation. Impact evaluations with smaller research samples tend to report fewer impacts. In these smaller samples, the difference may need to be substantial in order to pass tests of statistical significance. On the other hand, impact evaluations with larger research samples tend to report more policy impacts. Data used in impact evaluations Most impact studies use one of two types of data: administrative data or survey data. Administrative data are those used in administering Federal-State transfer programs. They are sometimes used for research because it is possible to economically collect large amounts of data on individuals over time without the response errors that occur if participants are asked about their public assistance receipt or work activities. However, administrative data come from data systems used to determine eligibility and benefits, not systems designed specifically for research. Therefore, evaluators are limited to the outcomes collected as well as the format in which these outcomes are collected. Welfare and food stamp receipt generally come from public assistance data files used to store information on a recipient household's characteristics, income, and benefit receipt. Employment and earnings information generally are from unemployment insurance (UI) wage files that track UI wages and taxes paid on behalf of an employee and are used to determine UI eligibility and benefits upon involuntary unemployment. These files do not collect information on all sources of employment, since the self-employed and Federal employees are not covered by the UI system. UI taxes and benefits are also based on quarters of earnings, and therefore UI wage files contain information only about quarterly earnings (not hourly wages, for example). Administrative data also do not capture information from outside the State. Therefore, if program participants move, their work and welfare activity are not going to be reflected in either the UI or public assistance administrative files. Researchers also use survey data to examine the effects of policy changes. Unlike administrative data, the primary purpose of survey data, which is collected through telephone surveys, mail surveys, or in-person interviews, is evaluation. Therefore, surveys are designed by those involved in the evaluation effort to solicit information on the outcomes of interest to the evaluation team. In addition, compared to administrative data, the costs of collecting survey data are higher. Because it is usually not possible to survey every individual in the targeted population, surveys often focus on a representative sample of the population, with the goal of generalizing these results to the larger group. For these results to be generalized, a large majority of the survey sample must participate. While evaluation studies that measure the impact of welfare policies on those who remain on the welfare rolls often obtain data on most participants, studies of former welfare recipients almost always have trouble with missing data because it is difficult to find participants once they leave the program. Another concern with survey data is that individuals may fail to understand the questions asked or respond truthfully and accurately. Interpreting impact evaluations The findings of impact evaluations are most valid for the particular place, time, context, and policy examined in the study. Generally, more confidence can be placed in findings that are replicated in a number of different settings. The findings of a random assignment evaluation are also limited to the population studied. This is often smaller than the entire population that might be affected by a policy. Most random assignment evaluations of welfare policy changes examine only those already receiving or applying for assistance. However, welfare policy changes might affect those not receiving assistance by making welfare more or less attractive to them. Random assignment impact evaluations generally do not measure these ``entry effects.'' In general, evaluation studies of welfare policy changes measure the effect of a package of initiatives (for example, a State's AFDC waiver program), rather than the effect of a single policy initiative. Researchers interested in the effect of a specific policy initiative have turned to a more sophisticated method of analysis, often introducing a third comparison group. In the Minnesota Family Investment Program (MFIP) and the Vermont Welfare Restructuring Project (WRP), for example, participants in a third comparison group (besides the regular experimental and control groups) were subject to the same policy initiatives as the experimental group, but were also subject to an additional policy initiative (for example, a requirement to work after 6 months of assistance). With the same random assignment and basic assumptions, any observed differences between the two experimental groups may be attributed to the additional policy initiative. Researchers have also used a third comparison group to evaluate an additional package of policy initiatives, an example being the National Evaluation of Welfare-to-Work Strategies (NEWWS). In the NEWWS design, the control group was subject to the former AFDC rules, one experimental group was subject to a package of work first initiatives and a second experimental group was subject to a package of basic education initiatives. Any observed differences between each of the experimental groups and the control group could be attributed to the different package of initiatives. The goal of this evaluation design was to measure the effect of a program that stressed work first versus a program that focused more specifically on education. The limitations of random assignment experiments have led some to use alternative methods to estimate the impact of policy changes. A few evaluations have used a ``quasi- experimental'' design. While the experimental design allows researchers to evaluate the program's impact, measured as the difference between the experimental and control groups, the quasi-experimental design allows researchers to evaluate the program's what-if effect. This is measured by comparing outcomes, after implementation of the policy changes, to those that would have been expected without the policy change, using administrative data as a baseline. Statistical techniques are used to ``control'' for various factors that could affect the examined outcomes (these factors would be controlled for in the experimental design through random assignment). While quasi- experimental designs use administrative data and therefore allow for a much larger sample, results can be sensitive to the specific technique used as they rely on many assumptions. Included studies The remainder of this section presents results from a review of impact evaluations of programs undertaken since passage of the Family Support Act of 1988 that serve the cash welfare population and, in a few cases, other low-income individuals. The discussion focuses on evaluations available as of June 2000. The discussion is organized into four sections: 1. Welfare to work, which focuses on findings related to earnings, employment, and receipt of cash assistance; 2. Family formation and structure, which focuses on findings related to marital status and childbearing; 3. Economic status, which focuses on findings related to the overall income and well-being of single mothers including health assistance and child care; and 4. Child well-being, which focuses on findings related to child development, including academic achievement, behavior, and health status. These categories were chosen based on the stated purposes of TANF, and on the types of outcomes evaluated in the studies. Each of the four sections begins with a background discussion of the relevant policy issues, highlights evaluations with noteworthy results, and provides a brief overview of evaluations that have reported information for outcomes relevant to that section. Since many of the evaluations examined impacts on several different outcomes (e.g., welfare to work and child well-being), the same evaluations are discussed in more than one of the four sections. Most of these evaluations are based on pre-TANF policies. However, their results have informed the creation of TANF and many of the policy initiatives tested have been included in State TANF Programs. States have broad flexibility to design their TANF Programs with minimal Federal requirements. Each TANF Program is unique and therefore, the results for one study may not be generalizable to a program in another State. Ideally, the results highlighted in this appendix should be interpreted in the context of the specific program's environment (such as the State's economy and population), in addition to the overall design and implementation of the program. Table L-2 describes major policies tested in these evaluations. Welfare-to-Work Impacts Promoting work to end dependence on government benefits is part of the stated purpose of the Temporary Assistance for Needy Families (TANF) Program. TANF requires recipients to engage in work (as defined by the State) within 2 years, and requires States to penalize recipients who fail to meet work requirements. In addition, States are subject to work participation standards. The effect of welfare programs on work has been the focus of a voluminous amount of research since the late 1960s. Studies that concentrate on the effects on employment, earnings, and continued welfare receipt constitute the bulk of the research undertaken since the 1988 enactment of the Family Support Act. TABLE L-2.--POLICY CHANGES UNDER WELFARE REFORM ------------------------------------------------------------------------ ------------------------------------------------------------------------ Asset limit increase...................... Set countable asset limit above the $1,000 of AFDC law and/or permit recipients to have a vehicle of greater value than allowed by AFDC law ($1,500 in equity value). Child support liberalization.............. Disregard more than $50 monthly of child support. Earnings disregard enhancement............ Disregard more earnings than AFDC law allowed (after 4 months of work, the AFDC disregard was $120 monthly). Education-focused programs................ Require participation in a program of education and training. Employment-focused program................ Require participation in an employment-focused program, such as job search. Family cap................................ Deny increase in family benefit (or pay less than full benefit) for new baby conceived or born to mother already on welfare. Food stamp benefit combined with cash Combine cash aid and food grant. stamps into one cash grant. Mixed employment/education program........ Assign recipients to either an employment-focused or education/training program. Personal responsibility sanctions......... Penalize family for failure to meet personal responsibility rules like assuring school attendance or taking the child for immunizations. Time limit................................ End family benefits when adult has received assistance for a specified period. Time limit (adult only)................... Reduce family benefits (end adult share) when adult has received assistance for a specified period. Transitional benefits extension........... Extend transitional Medicaid and/or child care beyond the 12 months required by AFDC law (for those who lose eligibility for cash aid due to increased earnings). Two-parent family eligibility............. End special eligibility rules for two-parent families. Wage supplementation...................... Use welfare benefit to subsidize wages of recipients. Work requirements: younger child.......... Require mothers with children younger than age 3 to participate in work/ training (AFDC required work if youngest child was 3, but permitted States to lower this to age 1). Work sanctions............................ Penalize work failure more severely than AFDC/Job Opportunities and Basic Skills (JOBS) law allowed (e.g., loss of adult benefit and use of protective payee). Work-trigger time limit................... Require work as a condition of continued eligibility for assistance after being on the rolls for a specified period. ------------------------------------------------------------------------ Generally, the post-Family Support Act research shows: 1. Mandatory welfare-to-work programs are associated with increased employment, and often also increased average earnings and decreased cash assistance payments. This includes welfare-to-work programs with a strong employment focus (``work first'') as well as programs that provide education. The magnitude of the impacts varies greatly among the programs evaluated. 2. Though both employment- and education-focused programs have succeeded in raising employment and earnings, their impacts show up at different times. ``Work first'' programs have immediate impacts, but they sometimes fade. Such programs often speed up the process of finding a job for those who would have worked anyway, and some who find a job subsequently lose it. Education-focused programs sometimes have delayed impacts on raising employment and earnings. Two of the most effective programs--California's Greater Avenues for Independence Program (GAIN) in Riverside during the 1980s and early 1990s, and the Portland, Oregon program of the 1990s--were mixed services programs that were ``work first'' for some participants, but provided basic education for those determined to need it before entering the labor force. 3. Some programs that combine mandated participation in employment-focused activities with enhanced disregards of earnings (a financial incentive to work) have been particularly effective in raising earnings and employment. Moreover, these impacts persist. However (especially in States with high case benefit levels), they sometimes increase cash assistance payments and months on welfare as the enhanced earnings disregard permits working people to be on welfare at higher levels of earnings. No increase in cash assistance was found in States with low benefit levels. 4. Most of the evidence from programs that included time limits (Connecticut, Florida, and Virginia) failed to show impacts on receipt of cash benefits before families reached the time limits. In other words, anticipation of benefit-termination time limits did not cause recipients to leave welfare early. However, all three programs also had enhanced earnings disregards that could have offset any behavioral response to the time limit. The disregards continue cash assistance eligibility for some whose earnings otherwise would exceed program limits. 5. The ``messages'' and expectations that a program conveys to participants affect its impact. Program messages, or the strength of the messages, often vary from site to site. For example, all California GAIN Programs in the late 1980s and early 1990s operated under the same statewide rules. However, the Riverside program projected a strong message that its purpose was for recipients to get employed quickly. Probably as a result of this message, the Riverside program also had some of the largest employment and earnings impacts of evaluated welfare-to-work programs. The program message might also affect how participants respond to time limits. Because of the variation in program messages, even among programs with the same rules, it is difficult to generalize evaluation findings from a specific site to other programs that might nominally have similar rules. Background and analytical issues Research shows that offering cash welfare to families reduces work effort (Danziger, et al., 1981). The decision to work requires a choice between the benefits that can be obtained in the work force and those available from allocating time to domestic activities (such as childrearing) or leisure. Though the low-income population is diverse, the low wages that are offered to most make welfare an economically more attractive option than a job. Additionally, some who are working might decide to reduce their hours of work and make up part of their lost earnings with the welfare payment. The major issues in welfare-to-work evaluations can be classified into two groups: 1. Can programs be designed to overcome the work disincentives inherent in welfare? Specifically, can work efforts be increased by lowering the rate at which earnings cause a cut in benefits, or, can these disincentives only be overcome by requiring mandatory participation in a work or job preparation program? 2. Can programs be offered that increase the earnings capacities of welfare recipients so that work becomes more attractive than welfare? What type of programs are most effective, ``work first'' or education-focused? \4\ --------------------------------------------------------------------------- \4\ For an overview of theoretical issues of the labor supply of women, see Killingsworth & Heckman (1986). See Becker (1965) for a discussion of theory of the allocation of time between market activities (work) and domestic activities. --------------------------------------------------------------------------- Work incentives/disincentives.--Empirical research dating back to the 1960s provides evidence that welfare is a disincentive to work (Danziger et al., 1981; Moffitt, 1992). To counteract the work disincentives, policies have been sought to either lessen the disincentive or require some participation in work or job preparation activities. One set of policies seeks to ``make work pay,'' often by reducing the rate at which a family loses welfare benefits as their earnings increase. Under the former AFDC Program, after 4 months on a job, cash benefits were reduced $1 for each additional $1 in earnings above $120 a month. Thus, a family's total income did not change with increased work effort, as it faced an implicit marginal tax rate of 100 percent for earnings above $120 monthly. Increasing the amount of earnings disregarded in calculating welfare benefits lowers the implicit tax rate faced by recipient families. However, the bulk of empirical research on the effectiveness of work incentives, for both increasing work and reducing welfare receipt, was discouraging (Moffitt, 1992). Increasing earnings disregards raises the level of income at which a family loses eligibility for benefits. This keeps more families on the rolls and could actually decrease work, through reductions in work effort that are at least partially offset by the welfare payment. The welfare caseload could increase, rather than decrease, as some who were previously ineligible for benefits would be made eligible by increasing earnings disregards. Nonexperimental research indicates that any increase in work effort by those already on the rolls would be canceled out by reduced work effort among new recipients. The disappointing findings of research on work incentives during the 1970s and 1980s may have played a role in policymakers' decisions to consider programs of mandatory participation in work activities or job preparation. Work incentives were sharply limited by the Omnibus Budget Reconciliation Act of 1981 (Public Law 97-35), which also permitted States to require welfare recipients to work in exchange for their benefits. Mandatory welfare-to-work programs impose a ``time cost'' on welfare recipients, which reduces their ability to allocate more time to domestic activities or leisure. For some programs (e.g., those that require work in exchange for welfare), the imposition of this time cost makes welfare less attractive and theoretically should reduce welfare participation. However, if these activities also provide benefits to recipients (such as education and training), the time costs might not result in reduced participation. Enhancing earnings capacities.--Higher earnings are associated both with higher levels of formal education and training and job experience (Willis, 1986).\5\ Though the welfare population is quite diverse, the most disadvantaged recipients tend to have relatively low levels of schooling as well as relatively little work experience. Higher earnings capacities can be expected to make work more attractive than welfare. However, there has been an ongoing debate about which of the following policies are more likely to move and keep people off welfare by enhancing their earnings capacity: --------------------------------------------------------------------------- \5\ Earnings and earnings capacities of individuals are usually analyzed using ``human capital theory,'' which relates wages to the skills individuals bring to a job (Becker, 1964). The most common empirical model of wages was developed by Mincer and Polachek (1974), with earnings estimated as a function of schooling and job experience. For estimates of earnings capacities for single women with children, see Haveman and Buron (1993). --------------------------------------------------------------------------- --programs that quickly move recipients into work to acquire work experience, often by forgoing formal education and training, with job search the most common first activity required, --programs that encourage recipients to acquire education and skills, often by deferring work and acquisition of job experience, or --programs that provide a mix of services, with the first activity determined by an assessment of individual skills and needs. Highlights of selected evaluations The post-Family Support Act welfare reform programs experimented with a wide variety of welfare-to-work strategies. In general, where evaluations compared participants assigned to a mandatory welfare-to-work program with recipients not assigned to the program, those in the welfare-to-work program had higher employment rates, higher average earnings, and lower average welfare benefits. Average earnings impacts are calculated based on the earnings of all members of an experimental or control group, including nonworking members who have $0 earnings. Average earnings have been increased by: (1) increasing employment; (2) increasing hourly wages of those employed; or (3) a mix of both. The followup period for these evaluations varies. For evaluations with a followup period of 2 years or less, impacts reflect short-term effects of a program. Mandatory welfare-to-work programs.--Almost all of the welfare-to-work programs evaluated since passage of the Family Support Act are mandatory work or job preparation programs. That is, welfare recipients were required to participate in activities or face a financial sanction. The degree of the financial sanction varied from program to program. Additionally, the degree to which participation requirements were enforced varied from program to program, but most were at least nominally mandatory. The Job Opportunities and Basic Skills (JOBS) Program created by the Family Support Act of 1988 mandated participation of most able-bodied Aid to Families with Dependent Children (AFDC) parents whose youngest child was age 3 or older in a program that provided a wide array of employment and education and training services. States had the option of mandating participation for parents with a child as young as age 1. Services were provided subject to available funding. The purpose of JOBS was to assure that needy families with children obtained education, training, and employment services to help them avoid long-term welfare dependence. There were two major evaluations of the JOBS Programs--an evaluation of the Nation's largest JOBS Program, California's GAIN (Riccio et al., 1994); and NEWWS (U.S. Department, 2000a). Both evaluations were fielded in multiple sites, and tested a number of different approaches to moving welfare recipients into the work force. GAIN was a single program, but was implemented differently in different sites. NEWWS is an evaluation of 11 different welfare-to-work programs. In the GAIN and NEWWS evaluations, recipients assigned to the program group were subject to participation requirements and were provided services; those assigned to the control group were not subject to participation requirements or provided services, but could otherwise seek employment and training in their communities. These evaluations provide information about the impact of welfare programs operating in the early-to-mid 1990s, before widespread use of AFDC waivers and before the creation of TANF. Over a 3-year followup period, the GAIN Program produced employment gains in five out of six sites, earnings gains in three out of the six sites, and reductions in cash welfare payments in four out of the six sites. The GAIN Program provided basic education for those assessed in need of education (i.e., they lacked a high school diploma or basic skills) and job search for others. However, county-by-county implementation of GAIN varied, with some counties emphasizing education more than others. The impacts in Riverside were particularly noteworthy; the employment rate was increased by 26 percent and average total earnings were increased by 49 percent over the 3-year followup period. The Riverside program was characterized by a strong employment message provided to all, even those assigned to basic education under GAIN's rules. Impacts from 2 years of followup data are available for all 11 programs in the NEWWS evaluation. NEWWS tested programs that had an employment-focused approach as well as education-focused programs. The 11 programs are: Atlanta, Georgia (employment- focused); Atlanta (education-focused); Grand Rapids, Michigan (employment-focused); Grand Rapids (education-focused); Riverside, California (employment-focused); Riverside (education-focused); Columbus, Ohio (integrated case management); Columbus Ohio (traditional case management); Detroit, Michigan; Oklahoma City (applicants only); and Portland, Oregon. Statistically significant increases in the percent ever employed in year 1 or 2 were found in all programs except for Oklahoma City (a program that evaluated impacts on applicants only) and the two Columbus, Ohio programs (that tested integrated and traditional case management). However, by the last quarter of year 2, significant impacts were found for the two Columbus programs, so that the evaluation found increases in employment for all programs except Oklahoma City. The NEWWS evaluation provides a comparison of programs emphasizing job search with those emphasizing basic education. Participants without a high school diploma or general education degree (GED) were less likely to enroll in basic education without mandated participation in a welfare-to-work program. However, education-focused programs had small impacts on participation in vocational and postsecondary education. This is due, in part, to many participants with a high school diploma or GED in both the experimental and control groups enrolling in these activities on their own. Evaluators also noted some other barriers to vocational and higher education, such as failing to meet entrance requirements (other than a high school degree) or the timing of required program activities failing to coincide with the timing of an academic year or semester. Comparison of mandatory programs: ``work first'' versus education and training.--Though NEWWS found positive impacts across both ``work first'' and education-focused programs, there were differences in both the degree and timing of these impacts. Chart L-13 shows the employment impacts of the 11 NEWWS programs over the 2-year followup period. The impact is expressed as the percent increase or decrease in the program group's employment rate, compared with the control group's employment rate. The chart shows the overall impact (percent change in the percent ever employed over the 2 years), as well as the impact in the last quarter of the second year. As shown in the chart, employment-focused programs had somewhat larger percentage impacts, though it was common for the impact to be less at the end of the 2-year period than over the entire period. Education-focused programs, on the other hand, tended to have their greatest impact at the end of the 2-year period. The fading of program impacts in some ``work first'' programs reflects a combination of two factors. Some who find a job subsequently lose it. In addition, some control group members eventually find jobs on their own without the aid of services provided to those in the experimental group. That is, some of the impact of employment-focused programs is simply an acceleration of the job finding process for those who would have gone to work anyway. The Portland, Oregon program is notable for relatively high overall impacts (18 percent increase in employment and 35 percent increase in average earnings over the 2-year period), as well as relatively high impacts at the end of the 2-year period. Portland represents a mixed services program, much like the Riverside program included in the GAIN evaluation. Both programs emphasized employment as their goal. However, both programs also directed some participants to basic education if they were determined to need skill building before entering the work force. Though the importance of job finding was the goal of both programs, the messages conveyed to participants differed. The Riverside message was that recipients should try to obtain any job, even one with low pay and no benefits as a ``stepping stone'' to a better job in the future. The Portland message was that recipients should try to find ``good jobs,'' those that pay above the minimum wage and provide benefits, and that it was OK to turn down some jobs. Despite similarities in earnings impacts, the Riverside GAIN and Portland NEWWS programs produced these impacts differently. Increases in employment contributed to both programs' impacts, but Portland also increased average hourly pay for those who worked. In Riverside GAIN, employed members of the experimental group had slightly lower average hourly pay than employed members of the control group (U.S. Department, 1998). CHART L-13. NATIONAL EVALUATION OF WELFARE-TO-WORK STRATEGIES: IMPACTS ON EMPLOYMENT RATE * Denotes impact that is not statistically significant. Note._LFA on the chart denotes labor force attachment, employment-focused programs; HCD refers to human capital development, the education-focused programs. Source: Prepared by the Congressional Research Service based on information from the U.S. Department of Health and Human Services and the U.S. Department of Education (2000a). The best test, among available evaluations, of whether employment-focused or education-focused programs are more effective can be found in the NEWWS evaluations in Atlanta, Grand Rapids, and Riverside. As mentioned, the NEWWS evaluation had an experimental group subject to an employment-focused program and an experimental group subject to an education- focused program. Both treatment groups were measured against the same control group. Therefore, the evaluators were able to measure the impact of an employment-focused program, the impact of an education-focused program, and the impact of the two approaches in comparison with each other. Table L-3 shows the comparison of the impacts of the employment-focused and education-focused programs. Generally, both types of programs increased earnings. Exceptions were education-focused programs for those without a high school diploma or GED, and the employment-focused program in Grand Rapids for those who had a high school diploma, though impacts for the programs themselves are not shown in the table. However, when differences between the two approaches were found, employment-focused programs tended to increase earnings more than education-focused programs over the 2 years, especially among those without a high school diploma or GED. Additional years of followup data might change these results, since there is a pattern of declining impacts for employment- focused programs as well as increasing impacts for education and training programs. However, previous studies have found that employment and earnings impacts of education-focused programs sometimes also fade after 2 years, although one program (Baltimore Options) did produce earnings increases at the end of 5 years (Friedlander & Burtless, 1995). TABLE L-3.--NEWWS. COMPARISON OF EMPLOYMENT-FOCUSED WITH EDUCATION- FOCUSED PROGRAMS IN ATLANTA, GRAND RAPIDS, AND RIVERSIDE: IMPACT ON EARNINGS OVER 2 YEARS ------------------------------------------------------------------------ Comparison of employment- focused program and education-focused program ------------------------------------------------------------------------ Full sample Atlanta............................... Employment-focused program impact greater by $521. Grand Rapids.......................... No difference. No high school diploma or GED Atlanta............................... No difference. Grand Rapids.......................... Employment-focused program impact greater by $1,063. Riverside............................. Employment-focused program impact greater by $707. With high school diploma or GED Atlanta............................... No difference. Grand Rapids.......................... No difference. ------------------------------------------------------------------------ Source: Table prepared by the Congressional Research Service based on information from the U.S. Department of Health and Human Services and the U.S. Department of Education (December 1997). Comparison of mandatory welfare-to-work programs: AFDC waivers.--The findings from welfare-to-work experiments conducted in the mid 1990s under AFDC waivers are not as consistent as those from evaluations of JOBS Programs. Evaluations of waiver programs compare participants subject to the welfare reform initiatives with those subject to regular AFDC/JOBS rules. Unlike the evaluations of JOBS Programs, which compared participants in a welfare-to-work program with those not subject to welfare-to-work rules, the waiver evaluations measured the incremental effect of new program rules above and beyond the effects of AFDC/JOBS requirements. A number of the waiver evaluations (Arizona, New Jersey, Texas) failed to find consistent impacts on employment, earnings, and welfare receipt. However, some waiver programs did have considerable effects; they increased employment and earnings and reduced welfare receipt (Delaware and Indiana). Delaware's ABC Program increased the employment rate of participants by 20 percent and increased earnings by 16 percent in the first year. Delaware also reduced cash assistance receipt by 5 percent in the first year. Indiana increased employment by 2 percent over 2 years and earnings by 5 percent. Indiana's program cut cash assistance payments by 20 percent, with impacts growing in the second year. Another set of waiver programs in a number of States used enhanced earnings disregards as an inducement to work. These programs are discussed in the next section. Combining mandatory participation and financial incentives.--Under waivers of AFDC requirements, States were permitted to experiment with higher earnings disregards than were set forth in Federal law. These programs combined financial incentives with mandatory work participation requirements, and their evaluation results are useful for anticipating the effects of changes made under TANF. Despite the disappointing evidence of earlier research on the effect of increasing earnings disregards, most States have coupled higher earnings disregards with mandatory work requirements in their TANF Programs. Earnings disregards in selected evaluated programs.--Under AFDC rules, States disregarded $120 plus one-third of remaining earnings for the first 4 months on a job when calculating benefits. For the next 8 months, the disregard was reduced to $120, and thereafter (after 12 months) to $90, a standard work expense. Most States under their TANF Programs have increased their disregards from those that prevailed under AFDC. --Connecticut: 100 percent of earnings up to the Federal poverty level. This applies to both cash assistance and food stamps. --Florida Family Transition Program (FTP): $200 and 50 percent of remaining earnings. --Minnesota MFIP: Benefits equal the maximum grant, increased by 20 percent minus net income. Net income includes an earnings disregard of 38 percent of gross earnings. MFIP consolidated cash assistance and food stamps into a single benefit. --Vermont WRP: $150 plus 25 percent of remaining earnings. --Virginia Initiative for Employment Not Welfare (VIEW): 100 percent of earnings until the cash benefit plus net earnings equal the Federal poverty level. The combination of financial incentives and mandated participation in work or job preparation has produced higher employment and earnings. Financial incentives alone, without a participation mandate, tend to produce smaller or no impacts on employment and often no impacts on earnings. MFIP and Connecticut's Jobs First, both programs that combined financial incentives and mandated participation, produced fairly large employment impacts. The final report on MFIP showed a 35 percent increase in employment and a 23 percent increase in earnings among long-term, single-parent recipients 9 quarters after the program began. Connecticut's Jobs First evaluation showed a 10 percent increase in the employment rate and an 11 percent increase in earnings over the 30-month followup. The impacts of both programs were sustained over time. Additionally, both programs had large impacts on the most disadvantaged participants, those without a high school degree, without work experience, and with long spells on welfare. In the case of MFIP and Connecticut's Jobs First Program, increases in employment and earnings were accompanied by increases, not reductions, in welfare payments and receipt of assistance. In these programs, the enhanced earnings disregards permitted some working recipients who would have been ineligible for AFDC to combine work and welfare. These programs could also cause a reduction in work effort or earnings, as those who would have worked anyway accept lower paying jobs or work fewer hours than would otherwise be the case, making up the difference in lost income with the increased welfare payment received while working. Whether increases in work effort by those already on the rolls are canceled out by reduced work effort by those not previously eligible for benefits (as found in earlier nonexperimental research) cannot be told from these evaluations. The findings of these evaluations are limited to the population studied: those who already applied for and/or received cash assistance. Thus, these evaluations cannot be used to understand ``entry'' effects; i.e., how programs affect the likelihood of low-income people to apply for welfare in the first place. Some evaluations of State programs that combined enhanced earnings disregards with mandated participation yielded different results than those found in Minnesota and Connecticut. Increased cash assistance payments were not found in the Florida FTP, the Vermont Welfare Restructuring Project (WRP), or the Virginia VIEW Program. Florida and Virginia pay lower benefits than do Minnesota and Connecticut, and the Vermont earnings disregard is less generous than those found in Minnesota or Connecticut. Further, in Connecticut and Minnesota, the cash and food stamp earnings disregards were both changed, while the Food Stamp Program in Florida, Vermont, and Virginia operated under Federal food stamp earnings disregard rules. Time limits: impacts before reaching the time limit.--Three evaluations are available on the possible effects of benefit- termination time limits on welfare-to-work outcomes: the Florida FTP Program, Connecticut's Jobs First Program, and the Virginia VIEW Program. The findings from the Florida, Connecticut, and Virginia programs are useful in anticipating the effects of TANF's time limit. All three programs combined time limits with other welfare reform initiatives, particularly enhanced earnings disregards that allowed some working participants who would have been ineligible for AFDC to continue receiving assistance and combine work and welfare. Most State TANF Programs are operating under policies that combine increased earnings disregards with time limits. To the degree that generous earnings disregards increase months of receipt for working recipients, they increase the risk of reaching the time limit, though they also reduce the risk of reaching the time limit without earnings. Generally, there is little evidence that anticipation of time limits spurred recipients to leave welfare early. No change in cash assistance receipt was recorded before participants reached the time limit in Florida or in two of the three sites evaluated in Virginia. Cash assistance increased in Connecticut. It is possible that the behavioral effect of time limits, which serve as a spur to leave welfare early, and the enhanced earnings disregards, which permit working recipients to stay on welfare longer, cancel each other out. In Connecticut, the effect of the very generous earnings disregard might have overwhelmed the effect of the short time limit. One site in Virginia (Petersburg) did show reduced cash assistance receipt after recipients became subject to time limits. In Petersburg, program staff actively encouraged working participants to leave the rolls, and save months of eligibility for future use. In other Virginia sites and in Connecticut and Florida, evaluators noted that staff did not actively encourage participants to save remaining months of eligibility. These results are further evidence that the program message conveyed by local administrators affects participant behavior. Welfare time limits have also had a high profile in the media and public discussions. Evaluations often found that many in the control group incorrectly thought they were subject to the time limit. In Connecticut, for example, 23 percent of the control group thought they were subject to a time limit. This misunderstanding by control group members would have the effect of reducing the impact of the time limit, since it is measured as the difference in outcomes between the program group and the control group. Programs for teenage parents.--Teen parents on welfare are at particular risk of becoming long-term recipients. The New Chance Demonstration, the Teenage Parent Demonstration (TPD), and Ohio's Learning, Earning, and Parenting (LEAP) Program all targeted teenage parents. New Chance was a voluntary program of education and services for AFDC mothers who had children as teenagers and dropped out of high school. The TPD required participation and provided enhanced services to teen parents. Neither set of programs produced lasting impacts on employment and earnings. New Chance slightly reduced employment and earnings early in the program, as participants engaged in education. The TPD produced some early impacts, but these did not last. Evaluators of the TPD found that impacts that occurred while young mothers were enrolled in the program were not sustained once the mandated participation requirements and expectations of the program ended for them. Ohio's LEAP Program imposed financial sanctions on those who failed to meet school attendance requirements, and provided bonuses for those who did attend school. Over the 4-year followup period, there were no employment and earnings impacts, but there was a 5 percent decrease in cash welfare. Detail for welfare-to-work impact evaluations.--For detailed information about the major impact evaluations conducted after the passage of the Family Support Act of 1988, see table L-9 at the end of this section. Family Formation and Structure Impacts Two of the objectives of TANF are to ``prevent and reduce the incidence of out-of-wedlock pregnancies'' and to ``encourage the formation and maintenance of two-parent families.'' Policy initiatives to promote these objectives have included imposing a family cap (paying no benefit or a reduced benefit for a new baby born to a mother already on welfare) and removing special eligibility requirements for two-parent families. However, the bulk of post-Family Support Act research has focused on broader efforts to alter welfare, especially by increasing work, and the impact of these efforts on earnings and employment (as discussed in the previous section). The impact of policy initiatives on nonmarital births and marriage has received much less attention. Available findings generally show: 1. The majority of evaluations of welfare-to-work programs that examine marriage rates and birth rates for single- parent families report no impacts. One evaluation of an AFDC waiver program (Minnesota's MFIP) did report an increase in marriage among single-parent families and a decrease in marriage breakup among two-parent families. 2. State AFDC waiver initiatives that included family cap policies have produced inconclusive results. While New Jersey reported a decline in birth rates for AFDC mothers, Arkansas reported no impacts on birth rates for AFDC mothers as a result of family cap policies. 3. The few programs that have targeted teenage parents have, in general, reported no impact on the number of births. Background and analytical issues Since the 1970s, the overall birth rate for married women has declined, but the number of nonmarital births has substantially increased (see appendix M). This increase in nonmarital births has been attributed to women delaying the decision to marry, but not the decision to have a child (Bachrach, 1998). Growth in the welfare caseload, to a historic peak in 1994, was driven in part by the increase in the number of single female-headed households. Since 1994, caseloads have declined at a largely unexpected rate (almost half by fiscal year 1999), but the number of single mothers has remained relatively constant (between 9.8 and 10 million each year). Therefore, while the increase in single mothers may explain much of the earlier increase in the welfare caseload, the decrease in the welfare caseload has not been accompanied by a similar decrease in the number of single mothers. Economic models of marriage decisions contend that single people marry if the marriage provides a benefit to both individuals. For couples to remain married, both individuals must continue to benefit from the union (Becker et al., 1977). Welfare programs provide a means for single mothers to live independently, and therefore, like employment, may serve as a disincentive to marriage. AFDC assisted single mothers and many believed AFDC and other welfare programs encouraged divorce or single parenthood by offering a financial alternative to marriage. The Family Support Act of 1988 required every State to operate Aid to Families with Dependent Children--Unemployed Parent (AFDC-UP) Programs, which provided assistance to eligible two-parent families but contained special eligibility requirements for such families. Specifically, the principal wage earner in the family was required to have a significant work history and was not allowed to work 100 hours or more per month. Many believed these eligibility requirements had a negative impact on decisions to marry or remain married. Although earlier nonexperimental research found no relationship between the presence of State AFDC-UP Programs and marital stability (Winkler, 1995). Under TANF, a significant number of States have removed these special two-parent eligibility requirements in their efforts to promote and support marriage. In addition to being a possible disincentive to marriage, welfare programs are seen by some as providing an incentive for nonmarital childbearing. In a purely economic sense, many argue that by increasing the cost of having children by limiting or denying benefits to children born to mothers already on welfare (the ``family cap''), public policy may be able to encourage people to postpone having a child or to limit family size. Others argue that by providing easier access to family planning services, and also by enhancing these services, public policy may decrease the cost of preventing unwanted or unplanned pregnancies, which may decrease the number of births. Still others argue that family size decisions, as well as decisions to marry or divorce, are subjective and personal and have little to do with economic calculations. The nonexperimental research examining the effects of welfare on marriage and fertility produced mixed results. Earlier studies led many to conclude that the welfare system had no effect on these outcomes. However, recently, a slight majority of studies have concluded that welfare has led to a decrease in marriage and an increase in childbearing (Moffitt, 1998). These studies use various methodologies, different sources of data, and explore different periods of time. Many contend that the findings reported are sensitive to the factors and methods used to examine this relationship. Highlights of selected evaluations The impact of policy initiatives on marriage and fertility has proven difficult to examine. As discussed above, marriage and childbearing decisions are highly personal, and it may be difficult to directly influence these decisions through policy. In contrast, many believe that certain policy changes may reduce welfare entry by encouraging persons to marry or delay decisions to have children. However, welfare evaluations only capture the effect of policy initiatives on those who already receive welfare. Therefore, these evaluations measure only the effect of policy changes on encouraging marriage among single women who are already receiving welfare, and promoting marital stability among married couples that also are already receiving welfare. Likewise, welfare evaluations only discern the effect of policy changes on subsequent births; those that occur after a woman has already been on welfare. Impact on marriage.--The Seattle-Denver Income Maintenance Experiment, conducted in the 1960s and 1970s, was the first experimental test of the impact of transfer programs on marriage. The analysis suggested that this program, which guaranteed a minimum income level, increased marital dissolution among white and black couples and increased rates of marriage or remarriage among Hispanic single-parent families. However, reanalysis of these findings has led many to question the impacts originally reported. Further, differences between policy changes and the program structure and environments make comparisons between the Seattle-Denver Income Maintenance Experiment and the more recent evaluations difficult. The Family Support Act of 1988 authorized a set of State experiments that liberalized eligibility for two-parent (unemployed) families by permitting them to work 100 or more hours monthly, without making other policy changes. By limiting other policy changes, many hoped that the evaluations of the three approved State demonstrations--California, Utah, Wisconsin--would provide insight into the impact of changes to the 100-hour rule. Reports on these demonstrations have been completed, but have not been released. The majority of the pre-TANF evaluations of ``work first'' and education-focused programs, which examine marriage among individuals already receiving welfare, report no impacts for single-parent families. One evaluation of an AFDC waiver program, the Minnesota Family Investment Program (MFIP), reported an impact on the percent of parents who were married 3 years after the program began. However, this impact varied depending on the population examined and the types of services individuals received, as illustrated in table L-4. TABLE L-4.--MINNESOTA FAMILY INVESTMENT PROGRAM: IMPACTS ON MARITAL STATUS 3 YEARS AFTER THE PROGRAM BEGAN ------------------------------------------------------------------------ Full MFIP Program (financial Single-parent families incentives plus mandatory participation) ------------------------------------------------------------------------ Long-term welfare recipients............ Increase in percent currently married. Recent applicants....................... No impact. ------------------------------------------------------------------------ Full program (financial Financial incentives incentives plus Effect of adding the Single-parent families: subgroups package mandatory mandatory participation participation) requirement Long-term welfare recipients: urban----Increase in percent------No impact..............--No impact.------------- counties. currently married. Recent applicants: urban counties.... No impact.............. No impact.............. No impact. Child outcome sample: single mothers Increase in percent Increase in percent No impact. with children ages 2-9 when program currently married. currently married. began. ---------------------------------------------------------------------------------------------------------------- Financial incentives package and elimination of special two- Two-parent families parent eligibility requirements Welfare recipients: urban counties......--Increase in percent currently- married. ------------------------------------------------------------------------ Source: Table prepared by the Congressional Research Service based on information in Miller et al. (2000). Overall, among long-term recipients in urban and rural counties, MFIP increased the percent who were married 3 years after the program began. The program had no impact on the percent of recent applicants who were currently married 3 years after the program began. The evaluation of MFIP was unique in its design. In addition to the control group, which was subject to the old AFDC rules, single-parent families in the research sample were assigned to one of two treatment groups. Those in the first treatment group, the financial incentives package group, were subject to changes including the enhanced disregard of earnings (financial incentives). Those in the second treatment group, the full program group, were subject to these same changes, but were also required to participate in MFIP's employment and training services once they had received assistance for 24 of the past 36 months. This three-way design allowed the researchers to disentangle the effects of the program's two major components--financial incentives and mandatory employment and training services. The previous welfare-to-work section reported that programs like MFIP, which combine mandatory participation in employment and training with enhanced earnings disregards, have been particularly effective in raising earnings and employment among single mothers who are long-term welfare recipients. However, among long-term, single-parent recipients in urban counties, this combined approach in Minnesota produced no impact on marriage 3 years after the program began. As table L-4 illustrates, while there was an increase in the percent of recipients who were currently married among those subject to the financial incentives package, there was no impact for those who were subject to the full package of MFIP benefits. Again, there were no impacts on marriage reported for applicants in urban counties assigned to either of the treatment groups. The MFIP evaluation also included a survey of mothers who were single and had children between the ages of 2 and 9 when they entered MFIP. Among these women, the combined program increased the percent who were married 3 years later. As table L-4 illustrates, this increase is entirely a result of the financial incentives, as the addition of the mandatory participation requirement had no impact on the percent who were currently married. While these results are difficult to reconcile, they provide some evidence that financial incentives, rather than mandatory employment and training services, may have an impact on marriage among single mothers. Two-parent families in MFIP were subject to different provisions than single-parent families, including the financial incentives package, a requirement that one parent work or participate in employment-related activities when the family had received assistance for 6 of the past 12 months, and the removal of special two-parent eligibility requirements. Many believed the removal of these eligibility requirements would promote marital stability by allowing two-parent families to remain together through difficult times such as loss of work. While the research design does not disentangle the effect of this specific reform