Manhattan Institute for Policy Research.
Subscribe   Subscribe   MI on Facebook Find us on Twitter Find us on Instagram      

Civic Report
No. 17 July 2001

Gaining Ground? Measuring the Impact of Welfare Reform on Welfare and Work

June E. O’Neill
Department of Economics and Finance
and Center for the Study of Business and Government,
Baruch College, CUNY

M. Anne Hill
Department of Economics, Queens College, CUNY
and Center for the Study of Business and Government,
Baruch College, CUNY


The report’s main findings are as follows:

  • The number of families on welfare declined by 50 percent between the passage of welfare reform legislation in August, 1996 and the date for the most recent caseload statistics, September, 2000.
  • Most of the women heading these families have gone to work, contrary to the expectations of many welfare reform critics. The proportion of single mothers who work has increased dramatically since welfare reform, nearly matching the proportion leaving welfare.
  • Regression results indicate that Temporary Assistance to Needy Families (TANF), the federal program created in 1996 pursuant to the welfare reform law, accounts for more than half of the decline in welfare participation and more than 60 percent of the rise in employment among single mothers.
  • These results also show that although the booming economy of the late 1990s contributed both to the decline in welfare and to the rise in work participation among single mothers, that contribution was relatively minor compared to the contribution of TANF, accounting for less than 20 percent of either change.
  • The decline in welfare participation was largest for groups of single mothers commonly thought to be the most disadvantaged: young (18–29) mothers, mothers with children under seven years of age, high school dropouts, black and Hispanic single mothers, and those who have never been married.
  • Employment gains have also been the largest among disadvantaged single mothers: mothers who have never married, mothers between the ages of 18 and 29, mothers with children under seven years of age, high school dropouts, and black and Hispanic mothers.
  • TANF’s beneficial effects extend even to the most disadvantaged portions of the welfare-eligible population. TANF accounts for 40 percent of the increase in work participation among single mothers who are high school dropouts; 71 percent of the increase in work participation among 18–29 year old single mothers; and 83 percent of the increase in work participation among black single mothers.



The history of welfare reform in the United States has recorded many failures and few successes. To the surprise of many observers, however, we now appear to be witnessing a major policy success. Between January of 1994 and September of 2000, eight and a half million people left the U.S. welfare rolls and the proportion of the total population on welfare declined from 5.5 percent to 2.1 percent, a level not seen since the early 1960s. Equally important, during the same time period single mothers—the dominant group of welfare beneficiaries—greatly increased their work participation as their welfare participation declined. Thus, the past six years have seen dramatic progress.

These dramatic changes in welfare and work occurred during a time of radical change in welfare policy itself. The most notable feature of that change was the passage in August, 1996, of the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA), which terminated Aid to Families with Dependent Children (AFDC), the nation’s basic welfare program since the 1930s, and replaced it with Temporary Assistance to Needy Families (TANF). TANF is a striking departure from the past. Under AFDC an eligible family was entitled to an income from welfare as long as it had a child under the age of 18 present in the home. Under the new law, welfare is no longer an entitlement either in terms of its financing or its philosophy. As its name suggests, TANF is intended to serve as a temporary helping hand. TANF benefits are restricted to a lifetime limit of five years, and all adult recipients must fulfill a strict work requirement.

PRWORA did not have an easy passage. Many welfare rights groups bitterly contested the “tough love” character of the policy reforms which had been written into the legislation by the Republican-dominated Congress. Three high level officials in the Clinton Administration resigned in protest. Nonetheless, President Clinton eventually signed the bill into law. Critics expected the worst. Writing in the New Republic, Katha Pollitt said,” . . . we know how welfare reform will turn out, too: wages will go down, families will fracture, millions of children will be more miserable than ever.”[1

The apparent failure of these predictions to come true has not stopped welfare reform critics. As increasingly positive news has been reported, critics of PRWORA have faced off with reform proponents over what is responsible for the record-breaking decline in the welfare rolls. Critics have alleged that most of the credit is due to the booming economy of the late 1990s, while reform proponents have contended that the law itself was the primary cause of these positive developments. Some critics concede that the law has reduced welfare rolls but claim that the result has been the impoverishment of families. Proponents point to the dramatic increase in work participation among those formerly receiving welfare. As of yet, however, there has been little definitive evidence to resolve the debate.[2

This report is the first in a series documenting the actual results of welfare reform. Here we examine in detail the changes that have occurred in the welfare and work participation of single mothers. We ask whether single mothers with the greatest socio-economic disadvantages have left welfare and entered the labor force to the same degree as mothers with more education and skills. Equally as important, we use regression analysis to estimate the extent to which welfare reform efforts themselves can explain the changes in welfare and work, taking account of the booming economy and many other relevant factors. Our main conclusions are as follows:

  • Contrary to the concerns of many critics, women who are thought to be least able to become self-sufficient have left the welfare rolls in large numbers since the mid 1990s. Single minority mothers who never married and those who dropped out of high school were among those with the largest absolute declines in welfare participation.
  • Single mothers entered the work force nearly as quickly as they have left the welfare rolls. Large increases in work participation were made by those with educational and other disadvantages, mirroring the pattern of change in welfare participation.
  • TANF is the most important single factor accounting for the decline in welfare and the rise in work participation among single mothers in the years since TANF was implemented. Regression results show that TANF accounts for half of the decline in welfare participation and more than 60 percent of the rise in employment among single mothers. The decline in unemployment—which we use as a proxy for the booming economy—during the TANF period accounts for less than 20 percent of either the decline in welfare or the increase in work.

These results strongly suggest that the positive news of the last few years is primarily the result of the passage of PRWORA.


The welfare caseload grew sharply during the late 1960s and early 1970s. This surge was sparked during the period when the so-called War on Poverty was launched and welfare benefits were enhanced and eligibility liberalized. The main policy response to the caseload growth was the provision of employment and training programs for welfare recipients. In addition, new financial incentives to work were provided by “disregarding” some of a recipient’s earnings in determining the reduction in her benefit amount. As a result of this provision, a welfare recipient could keep a larger share of her welfare benefits as her earnings increased and still remain on welfare.

By the 1980s it became clear that these policies were not working very well. Indeed, some observers concluded that the earnings disregard provision had contributed to caseload growth by making welfare participation more attractive.[3]  The disregard provision was eliminated by the Omnibus Budget Reconciliation Act of 1981 (OBRA), which also put a monetary cap on the amount of income a person could receive and still remain on AFDC. These restrictive eligibility provisions served to limit welfare caseloads through the 1980s.[4]  But despite a growing economy and rapid declines in the unemployment rate, welfare caseloads in virtually all states remained at or near their record-high levels of the late 1970s.

This fact sparked yet another attempt to reduce the welfare rolls, through the Family Support Act of 1988. Using the same approach of offering employment and training services, it established a Job Opportunities and Basic Skills Training program (JOBS) which added an educational component to the standard job training elements and provided transitional Medicaid support and childcare for recipients going off AFDC. As with the earlier earnings disregard program, however, this service enrichment may have done more to increase welfare participation than to reduce it. The welfare caseload rose more sharply between 1989 and 1994 than can be explained by the economic downturn of the early 90s. Part of that unexplained rise may be due to the welfare enhancement aspects of the JOBS program.

The rise in the AFDC caseload in the early 1990s and mounting criticism of the program from policy analysts and legislators (including Bill Clinton and Newt Gingrich) finally spurred reform measures that truly were departures from the past. Initially, individual states took the lead by conducting experiments and introducing more permanent reforms under waivers from the prevailing law granted by the federal government. During the period of waiver activity, 1992–1997, thirty states implemented a major waiver (Appendix A, Table A-1). In contrast to the earlier welfare reform efforts, the new state initiatives began to change the psychology of the program. A number of states introduced measures that restricted eligibility for benefits and imposed tough conditions for their receipt. Ultimately, the Congress passed PRWORA, abolishing AFDC and creating the TANF program.

Although TANF gives the states broad authority to determine conditions of eligibility for benefits, certain conditions are mandatory for all states—and these conditions shared a “tough love” philosophy similar to that found in many state waiver initiatives. Of particular importance is the imposition of the 5-year lifetime limit for receipt of TANF benefits and the requirement that all adult recipients fulfill a work requirement after a maximum of two years on the program. States have the option to impose tougher restrictions and many of them have done so (See Appendix A, Table A-2). For example, 19 states have a time limit stricter than 60 months, either because the limit is simply shorter or because it is shorter within fixed periods (such as no more than 24 months in a 60-month period). Many states also imposed strong work provisions; for example, requiring work registration immediately on entering the program, and job search and work activity well before 24 months. The use of sanctions and the policy of allowing exemptions from work only to those with very young children (under six months of age) have further toughened the work requirement.

On the other hand, states have some room to ease the TANF requirements. For example, they may choose to pay for benefits beyond the 60 month time limit with their own funds and are allowed to use federal funds for cases with particular handicaps, but are limited to 20 percent of the caseload for that purpose (See Appendix A, Table A-2).

This dramatic change in philosophy towards welfare sparked outcry and predictions of desolation, despair, and doom for the women and families affected. It has now been almost five years since PRWORA’s passage; what do the data show about its effects?


Overall the data are very encouraging. As shown in Figure 1, the welfare rolls have plummeted by more than half since late 1994, with most of that decline occurring after TANF’s enactment. Furthermore, as shown in Figure 2, the women who left welfare have not been left destitute; instead, they have joined the workforce in record-breaking numbers. Nevertheless, aggregate figures cannot tell the whole story. It is possible, as many critics allege, that those who have left welfare for work are primarily more educated, white women with work experience, leaving behind a caseload that is increasingly dominated by women with educational and other disadvantages. Another allegation is that when disadvantaged women—high school dropouts, black and Hispanic women, young mothers with young children—are pushed off welfare by work requirements and sanctions, they will not be able to find jobs and will consequently fall into extreme poverty. However, careful consideration of both the conceptual issues involved and the evidence indicate that these pessimistic forecasts have not materialized. Single mothers with disadvantaged characteristics sharply reduced their welfare participation in the years of welfare reform and just as sharply increased their participation in the work force.

A. Conceptual Issues

Under the old AFDC program, women with relatively little education and other disadvantages were more likely than better-situated women to go on welfare in the first place, and were more likely to stay on welfare for long periods if they did go on the program.[5]  There are obvious reasons for this differential. The welfare benefit package—cash benefits, food stamps, Medicaid, housing subsidies—is more financially attractive when the income alternatives to welfare are low, and disadvantaged women are more likely to have low potential earnings and poor marriage prospects.

Welfare reform, however, has greatly changed the relative attractiveness of welfare. The time limit ultimately eliminates the choice of being on welfare; and work requirements and tough sanctions change life while on welfare, reducing leisure time and time at home generally. Although these changes affect everyone, they represent the most dramatic change for those who would have been long-term recipients—accumulating more than five years on welfare—under the old system. Under that system, high school dropouts and those with an early first birth out of wedlock were particularly likely to be long-term welfare recipients.[6]  Potential long-term recipients would have a stronger incentive than potential short-term recipients to postpone entry to save up the five-year welfare allotment for a rainy day. Some may be shocked into rethinking their life situation and follow a different path—stay in school longer; otherwise acquire more work skills; postpone a first birth.

Because the change in policy is more radical for those with a higher propensity to be on welfare, the disadvantaged are more strongly affected by the change. However, single mothers with more education and fewer disadvantages may respond more quickly to policy changes because they are more informed and more capable of making the transition from welfare to work.

In this section we examine in detail what has actually happened and report on the trends in welfare and work participation of single mothers with different characteristics. This descriptive analysis provides a picture of the timing of changes in welfare and work participation and the extent to which that timing has been coincident with the implementation of welfare reform during the waiver and TANF periods. Comparing the trends in welfare and work for groups of single mothers who differ by race, schooling, age, and other characteristics related to the propensity of the group to go on welfare provides additional insight into the possible effects of welfare reform.

The examination of trends, while suggestive, cannot be regarded as conclusive. We turn later to a more rigorous statistical evaluation in which we control for the effects of unemployment along with a large number of other factors that could also influence welfare and work participation.

B. Data Sources

The data source for both our descriptive and regression analyses is the annual March supplement to the U.S. Current Population Survey (CPS), a large, nationally representative survey of American households conducted monthly. We analyze data on individual respondents from each of the March supplements from 1983 through 2000. Our use of individual, or micro data, is one important methodological difference between our analysis and that of most other studies of the effects of welfare reform which have analyzed aggregate state caseload data. In contrast to caseload data, the CPS micro data have the great advantage of providing information on the welfare receipt, employment, and detailed demographic characteristics of each person, thereby enabling the analyst to control more precisely for such factors (Appendix B compares the CPS data on welfare recipients with the standard caseload data).

We limit the sample to single mothers ages 18 to 44—i.e., never-married or previously married mothers who have children under the age of 18. This is the group of women who are most likely to be eligible for welfare benefits and therefore potentially “at risk” for receipt of benefits. The pooled sample, which we later use for regression analysis, contains approximately 80,000 single mothers. Although a majority of single mothers head their own independent households, a significant minority—23 percent—are “sub-family” heads living with their children in the household of a parent or another related or unrelated adult. We include all single mothers, whether independent or heading sub-families. We start the analysis in 1983 because that is the first year for which the CPS provided the information needed to identify the single mothers of children living in multi-family households. We choose an early start date so that our sample can span several changes in the business cycle, allowing more accurate measurement of the economic effects.

C. Single Mothers, Especially Disadvantaged Mothers, Have Left the Welfare Rolls in Large Numbers

Single mothers with the lowest skills and other deterrents to labor market success have always had the highest rates of welfare participation. Yet these groups have experienced as large or larger declines in welfare participation in the late 1990s than more skilled groups. As noted, the structure of the policy reforms suggests that the impact could differ among population subgroups. Figures 3 through 7 depict trends in welfare participation for women in different categories classified by age, age of children, education, race/ethnicity, and marital status. The results are as follows:

Age of Mother and Age of Youngest Child. The decline in welfare participation has been largest for the youngest mothers (Figure 3).[7]  One reason for an age differential in response is that the imposition of a time limit is likely to mean less of a change for older women than for younger women. Under the AFDC program, benefits ended when the youngest child turned age 18. Older women are likely to have older children and a woman whose youngest child is 13 to 17 is not likely to regard a TANF time limit as very different from the old AFDC rule that terminates benefits when the youngest child reaches age 18.[8]  More evidence for this explanation is provided in Figure 4, which shows a considerably milder decline in welfare participation after 1993 for mothers whose youngest child is 13 to 17 than for mothers with younger children. As we have explained, the time limit combined with a work requirement that in many states allows no exemption for mothers of children older than six months (even younger in a number of states), plus a possible family cap, add up to a benefit package dramatically reduced from AFDC days.[9]  Under these circumstances a young mother might not apply for welfare, or if already on the program, she might leave welfare for a job, school, or marriage. With the five-year time clock ticking and a good job market, young mothers may also choose to bank their welfare participation for the future.

Schooling and Race. Single mothers with the lowest levels of education have poorer labor market prospects and they are significantly more likely to take up welfare. However, since 1992, high school dropouts, the most disadvantaged group, have experienced a particularly large decline in welfare participation (Figure 5). Black and Hispanic women have less schooling than white non-Hispanic women, which is one reason why their welfare participation rates are higher (Figure 6). Nevertheless, since 1992 the participation rates of black and Hispanic mothers have declined more than those of white non-Hispanic mothers and the gap in rates narrowed noticeably by 1999.

Marital Status. Single mothers who never married typically have less education and are much more likely to go on welfare than are single mothers who were previously married and have been divorced, separated, or widowed. Yet again, however, the data show that these disadvantaged women left welfare much more quickly between 1992 and 1999 than their previously married counterparts, significantly narrowing the difference between the two groups in welfare participation rates (Figure 7). Once again the group of single mothers most likely to be affected by the change in the welfare program experienced the greatest absolute reduction in welfare receipt.

D. Single Mothers, Especially Disadvantaged Mothers, Greatly Increased Their Work Participation

Single mothers have entered the workforce nearly as quickly as they have left the welfare rolls, resulting in an impressive increase in their work participation. The timing of the rise in work participation suggests that welfare reform could be an important factor motivating the increase in work effort as well as the decline in welfare participation.

Marital Status. Trends in employment participation differ considerably among women according to their marital status. The employment rate of married mothers increased gradually over the 1980s and 1990s, but the employment of the two groups of single mothers (never-married and previously married) noticeably turned upwards only in the welfare reform era (Figure 8).[10] Moreover, the upturn is much more pronounced for never-married mothers than it is for previously married mothers. If the decline in unemployment and/or other trends in the economy favoring women were the main factors prompting this rise in employment, we would expect to see similar patterns of change for all mothers 18–44, regardless of their marital status. The fact that the upturn is strongly related to the degree of welfare dependence of the group suggests that welfare reform is likely to have played an important role in the rise in work participation.[11] In 1992, 64 percent of currently married mothers were employed during the week prior to the March CPS survey, while only 44 percent of never-married mothers were so employed.  In the span of time between 1992 and 2000 this gap was closed, with 67 percent of never-married mothers and 67.5 percent of currently married mothers employed during the March 2000 reference week.

Generally speaking, the pattern of employment gains made by sub-groups of single mothers almost mirrors (in reverse) the pattern of welfare declines. Dramatic employment gains were made by disadvantaged women who traditionally have had the greatest welfare attachment and were therefore likely to be most strongly affected by welfare reform.

Age of Mother and Age of Youngest Child. Similar to the pattern of change in welfare participation, the sharpest employment gains have been among the youngest women (Figure 9). While only 38 percent of 18–24-year-old single mothers were employed in 1992, 60 percent were so employed in March, 2000. Among 25–29 year olds, employment rates rose from 50 percent to 72 percent. In stark contrast, among 40–44 year old single mothers, employment rates were already quite high in 1992 (73 percent) and rose only moderately (to 79 percent) by 2000.

A similar pattern emerges when we compare the trend in employment among single mothers sorted by age of youngest child. The gain in employment for single mothers whose youngest child was less than seven was huge—from 46 percent employed in 1992 to 70 percent in 2000 (Figure 10). On the other hand, the percent employed among single mothers whose youngest child was 13–17 fell somewhat (it was 77 percent in 1992 and 75 percent in 2000). As noted, this group would not be affected by the five-year time limit since welfare rules under AFDC had always terminated benefits when the youngest child reached age 18.

Schooling and Race. Single mothers with low levels of education and skill lagged behind during the 1970s and 1980s as other women with children greatly increased their work participation. Yet the employment rate of this group increased more rapidly than that of other education groups during the 1990s (Figure 11). Thus, among single mothers who were high school dropouts the percent working rose from 31 percent in 1992 to 50 percent in 2000.[12]  The employment rate of high school graduates rose from 60 to 74 percent, and for single mothers with at least some college level schooling, it increased from 74 to 84 percent.

White single mothers typically have had higher employment rates than black or Hispanic single mothers (Figure 12). However, strong employment gains by black and Hispanic women have narrowed this differential considerably. In 1992 only half of black single mothers were employed, but by March 2000 their employment rate had climbed to 69 percent. Gains for Hispanic women were nearly as large, rising from 47 percent employed in 1992 to 63 percent in 2000. Among white single mothers employment rates rose more moderately—from 65 percent in 1992 to 79 percent in 2000.

Summary on Changes in Welfare and Work. In sum, the welfare participation of single mothers generally declined and their work participation rose during the period of welfare reform (1992–2000), first during the waiver years and then after TANF. Moreover, especially large declines in welfare participation and increases in work are observed among women with characteristics associated with high rates of welfare receipt: young single mothers with young children, black and Hispanic women, high school dropouts, and mothers who never married. In this respect, the fears of critics have proved unfounded. But we have yet to answer an important question: was this startling change in behavior the result of the booming economy or of welfare reform?


The overall patterns of change in welfare and work participation, and the patterns of those disadvantaged sub-groups considered to be most vulnerable to changes in policy, suggest that welfare reform was a causal factor spurring these changes. As noted, however, the sharp decline in unemployment that occurred over the same period is also likely to have played a role. The true effect of welfare reform cannot be determined without accounting for changes in unemployment and other possible factors affecting single mothers’ choices.

A. Research Strategy

Our empirical analyses address two different questions. First, using regression analysis, we ask the causal question: Other things staying the same, what is the effect of the welfare policy changes brought by the state waivers and TANF on the probability that a single mother was on welfare? “Other things” include the level of unemployment and wage rates in the woman’s state of residence; the generosity of that state’s welfare program; and the woman’s relevant demographic characteristics (e.g., her education, her age, and the ages of her children). We conduct a similar analysis of the probability that a single mother is employed.

Second, we use the behavioral parameters to simulate (a) the share of the actual aggregate changes in welfare and work attributable to welfare reform, and (b) the share attributable to the economy.

B. Using Regression Analysis To Determine the Effect of TANF on Welfare Participation and Work Participation

To identify the effects of welfare reform, we rely on three kinds of comparisons.

First, because our data span the 18-year period 1983–2000 and major waivers were not introduced until 1992, we can compare the probability that a similarly situated woman would be on welfare (or would work) in years with and without a waiver or with and without TANF.

Second, similarly situated women in the same year were exposed to different welfare reform regimes, and that variation provides us with another basis of comparison. We take advantage of the fact that the 50 states and the District of Columbia varied in their use and date of implementation of waivers and also in their date of implementation of TANF.

Third, we conduct separate regressions for sub-groups of women—who differ by their propensity to be on welfare—and compare their responses to welfare-policy changes and to changes in the economy. Our comparison groups differ by education, by age, and by race.

The Variables. We use multiple regression analysis to analyze both welfare and employment outcomes.

The dependent variables are:

  • Whether or not a single mother received any public assistance income during the previous calendar year; and
  • Whether or not a single mother was employed during the March CPS reference week (the last week before the date of the survey interview).

Our choice of explanatory variables is intended to allow us to control for factors that would affect the welfare or work participation of single mothers in a given year and in a given state. The factors can be categorized as those that pertain to the individual’s personal demographic characteristics and those that relate to the economic conditions and welfare policies in the individual’s state of residence.

The explanatory variables are:

  • Waiver and TANF. Both measured as indicator or dummy variables, which turn on in a particular year for an individual woman when her state of residence implemented a major statewide waiver or TANF. If the policy was in effect for a full year the indicator gets a value of one. However, in the first year of implementation the policy typically was introduced during the year. In that case the value of the indicator is the fraction of the year that the policy was in effect (the indicator takes a value of zero in all years when no waiver or TANF is in effect) .
  • Unemployment. This factor is measured as a set of indicator variables denoting whether the individual lived in a state in a particular year in which the unemployment rate was (a) less than 4%, (b) 4–5%, (c) 5–6%, or (d) 6–7%, with 7% or over as the omitted, or reference variable. We chose this specification to capture the possible non-linear effects of changes in the unemployment rate. The data for these variables were obtained from supplementary tabulations of CPS data that measure the annual average unemployment in each state for each calendar year for use in the welfare participation models and the average unemployment rate in February and March for use with the employment participation models which refer, as noted, to outcomes in March.
  • Wage Rates and Welfare Benefits. Wage variables are intended to capture changes in the gain from work and from investments in work skills, and the maximum welfare benefit for a family of three is a measure of changes in the generosity of welfare benefits in each state and each year. The wage rate is estimated for workers with a high school education or less and an additional variable measures the ratio of the wage of college graduates to that of high school graduates.[13]  Wage and benefit variables are measured in constant dollars and in logarithmic form.
  • Demographics. Among these variables are:
    • Age of mother, specified as a series of indicator variables (18–21, 22–25, 26–29, 30–34, 35–39, with 40–44 as the reference variable).
    • Race of mother (with white non-Hispanic as the reference group and black non-Hispanic, Hispanic, or other non-Hispanic as the categories).
    • Education of mother (high school graduate as the reference group, and whether or not the mother was a high school dropout, had some college, or graduated from college as the categories).
    • Other: the number of children under 18; the age of youngest child (three categories—less than 7, 7–12, and age 13 or over—the last is the reference category); whether the mother was never married; whether or not the mother lived in the central city of an MSA or a non-MSA area, the reference group being residence in the ring—i.e., the suburbs and small towns immediately surrounding the central city.

Although we are able to control for a large number of factors that could affect the welfare or work participation of single mothers, we could not readily measure some factors which also might have an effect. Specific policies—increases in the value of the Earned Income Tax Credit, subsidies for childcare, Medicaid coverage for those off welfare, the cultural or political environment—can vary from state to state and over time. If they are correlated with our key variables of interest and are excluded from the analysis they could artificially enhance or obscure the true effects of these variables. In an effort to handle this potential problem we have followed standard practice and included (a) variables indicating the individual state of residence of each woman in each year, (b) a trend and trend squared variable, and (c) the interaction of each state variable with the trend variables.

Regression Results. After controlling for the state of the labor market and the other relevant variables, regression analyses enable us to estimate the impact of welfare reform on (a) the likelihood that a single mother receives welfare, and (b) the likelihood that she works. Our estimates—first for all single mothers by age group and then for education and racial sub-groups—are summarized in Tables 1, 2 and 3.

Differences by Age Group. Among all single mothers ages 18–44, the implementation of TANF had a powerful effect, reducing the probability of welfare participation by six and a half percentage points—an effect that is statistically significant as well as large in relative magnitude (Table 1).

Exposure to a state waiver had a weaker effect, however, reducing the probability of being on welfare by about half a percentage point. These effects are stronger among younger women. The weak measured effect of the state waivers is partly due to the small portion of the population that was initially affected by any waiver. Only about half of all single mothers lived in a state that had implemented a waiver by 1996. Moreover, the waivers varied considerably in their content—for example, less than half of those adopted included a time limit (Appendix Table A-1).[14]  TANF, which was implemented nationwide under federal law, required a time limit and a work requirement in all states. It is not surprising that the waivers, on the whole, have a weaker measured effect on welfare participation than TANF.

As anticipated, we also find that the greatly improved state of the labor market had a large and statistically significant effect on the welfare participation of single mothers. In a state where the unemployment rate fell below four percent, the welfare participation of single mothers living there was estimated to be six and a half percentage points lower than that of a similarly situated single mother in a state with an unemployment rate above seven percent.

The welfare participation of younger single mothers (ages 18–29) is considerably more sensitive to welfare policy than is the participation of older mothers (ages 30–44). Our estimates indicate that TANF reduced the probability of being on welfare among younger mothers by almost 13 percentage points, compared to 2.5 percentage points for older mothers.[15]  The response of younger mothers to waivers, although not nearly as large as it is to TANF, is nonetheless significant both in size and in a statistical sense. The effect of declining unemployment on welfare participation does not differ very much between older and younger single mothers.

Welfare policy had a similar effect on the work participation of single mothers as it did in reducing welfare participation. TANF again has a larger impact on welfare reduction than waivers, although the impact of waivers on the propensity to work is somewhat stronger than it was regarding the propensity to be on welfare. Thus, for women ages 18–44, the effect of waivers on welfare participation was weak and statistically insignificant while waivers had a small but statistically significant effect on work, producing a 2.3 percentage point increase in participation. TANF, however, has an equally strong effect on work as on welfare, increasing work participation by 6.6 percentage points and reducing welfare participation by 6.5 percentage points for all women ages 18–44. The effect of TANF on work participation is greater among younger women ages 18–29 than it is for women 30–44, increasing the employment rate of the younger group by 8.1 percentage points and that of the older group by 5.4 percentage points.

The effect of the state’s unemployment rate also bears a strong and statistically significant effect on work participation. Among all single mothers ages 18–44, living in a state with an unemployment rate of less than four percent is associated with an increase in work participation of 5.6 percentage points relative to similar women in states with unemployment rates exceeding seven percent.

Differences by Education and Race. TANF had a large and statistically significant effect on reducing the probability of welfare participation at all education levels (Table 2). However, the effect is largest for college-educated women. This finding initially may appear to be inconsistent with Figure 5, which graphically depicts a larger decline in welfare participation among high school dropouts since 1992. The graph, however, gives the aggregate picture of change due to many causes. Moreover, as we noted in the conceptual discussion, it is possible that educated women respond more quickly to policy changes because they are generally more informed and adaptive. Waivers had a much smaller but still statistically significant effect in reducing welfare participation among college-educated mothers. However, that effect was insignificant (and in the wrong direction) among mothers with less education.

The effect of TANF on work participation is both large and statistically significant among women at all levels of schooling, increasing the employment rate of high school dropouts by 6.3 percentage points, of high school graduates by 5.6 percentage points, and of those with some college by 7.6 percentage points.

Comparing the results by race and Hispanic origin (Table 3), we find that black single mothers experience the strongest effect of TANF on both the probability of welfare and the probability of work participation. TANF is associated with a very large and statistically significant reduction in welfare (8.4 percentage points) and an equally large 8.7 percentage point increase in work among black single mothers. TANF is also shown to have a strong effect on both the welfare and work participation of white non-Hispanic mothers. However, the response of Hispanic mothers to TANF appears to be much weaker than that of either black or white mothers. The low response of Hispanics is particularly surprising in view of the sharp restrictions placed by PRWORA on the eligibility of immigrants for TANF benefits, since Hispanics have a higher proportion of immigrants than other groups.

C. Welfare Reform Is Responsible for Over Half of the Decline in Welfare Participation Since 1996, and Over 60 Percent of the Rise in Work Participation

Our regression results simulate the contributions of both welfare reform and the economy—as measured by the decline in unemployment—to the actual aggregate changes in the welfare and work participation of single mothers during the 1990s. We have made separate estimates for the pre-TANF period, when the state waivers came into effect (1992–1996), and for the period when TANF was implemented (1996–1999).[16]  Our estimates are shown in Tables 4, 5, and 6.

The relative contribution of welfare reform is presented as a ratio of the predicted change induced by welfare reform (or the economy) to the actual total change in welfare (or work) participation. For example, the procedure for estimating the contribution of welfare reform to the decline in welfare participation during the waiver period first measures the change in the proportion of mothers exposed to a waiver, then multiplies that change by the general effect of waivers on welfare participation derived from the regressions (the coefficients shown in Table 1). That product is the predicted change in welfare participation attributable to waivers. To obtain the contribution of waivers, we express that predicted change as a percentage of the actual change in welfare participation. The contribution of TANF to the decline in welfare participation was estimated following the same procedure, but with the data relevant to TANF.

The inferred relative contributions vary in exact magnitude, depending on the population group. However, there are some clear patterns:

  • Welfare reform contributed the lion’s share to the decline in the welfare rolls during the post-TANF period, accounting for almost half of the 11 percentage point decline in welfare participation among single mothers. Declining unemployment accounted for less than 20 percent of the welfare decline (Table 4).
  • Welfare reform brought by waivers contributed less than economic factors to the decline in welfare participation in the pre-TANF period. While the decline in unemployment accounted for 30 percent of the decline in welfare participation in that period, the implementation of waivers accounted for only 12 percent.[17
  • Welfare reform accounted for an even larger share of the rise in work participation among single mothers during the post-TANF period (more than 60 percent) while, again, less than 20 percent could be attributed to declining unemployment. The contribution of declining unemployment to work participation was higher (35 percent) in the waiver period. During this period, welfare reform contributed 22 percent to the rise.
  • Welfare reform explains a much larger share of the decline in welfare participation among young single mothers than it does among older mothers. During the post-TANF period, welfare policy is estimated to account for more than 70 percent of the decline in welfare among younger mothers ages 18–29, compared to 29 percent among mothers ages 30–44. In both the waiver and TANF periods, the economy plays a relatively more important role among older mothers compared to younger ones.
  • Welfare reform accounts for an equally important share (one-half) of the decline in welfare participation among blacks, whites and Hispanics in the TANF period (Table 6). But reform accounts for an especially large share (83%) of the rise in employment among black single mothers.
  • In the TANF period, the contribution of falling unemployment to the decline in welfare participation is below 20 percent for all population sub-groups but one. The exception is older single mothers, for whom the contribution is only slightly higher (26 percent).
  • Patterns in the rise in work participation mirror patterns in the decline in welfare participation. During the waiver period, declining unemployment contributes somewhat more to the rise in work participation than it does to the reduction in welfare participation. During the TANF period, however, the decline in unemployment plays a minor role in increasing work participation. Thus, among all single mothers 18–44, welfare reform explains 62 percent of the rise in employment participation; falling unemployment explains only 17 percent (Table 4). Among black single mothers the welfare policy share rises to 83 percent in the TANF period, while declining unemployment accounts for only 21 percent.


The welfare reforms of the 1990s have altered the work and welfare behavior of single mothers. While some of the decline in welfare numbers is, indeed, attributable to the booming economy, welfare reform has been a more important factor—accounting for one-half or more of the changes.

The policy changes have also been successful in reducing welfare and increasing work among disadvantaged women. Contrary to the fears of critics that they would be left behind, in many instances they have made the largest changes. As detailed in Appendix C, the single mothers who are currently on welfare are still less advantaged than other women, although they are no more or less so than they were before welfare reform.[18]  However, the number of women on welfare is now much smaller than it was.

Will the gains that have been made be sustained? There are two kinds of risks. One concerns the economy; the other, politics. An economic downturn is bound to occur at some time; indeed, it may be occurring as this research is released. Using the results of our statistical model, we estimate that a downturn in the economy that increased unemployment to the level of 1993 would increase the welfare participation rate of single mothers by five percentage points. That higher rate is still considerably lower, by 15 percentage points, than the welfare participation rate that actually prevailed in 1993. However, that estimate assumes, among other things, that PRWORA will not be weakened by legislation and that states will continue to implement policy as strictly as today.

Of course, political outcomes are no more predictable than the economy. If the PRWORA reauthorization chips away at the time limits and weakens work requirements, welfare participation could mount once again.

We have reason to be optimistic, however. The combination of welfare reform and a tight labor market has enabled a very large number of single mothers to gain work experience. Earnings rise with work experience and this relation is as true for former welfare recipients as it is for others. As experience and earnings rise, the less likely a woman is to slide back to welfare. The changes will be reinforced if the new generations of young women postpone childbirth and stay in school longer. If these long-term changes do occur in the next few years, then it can truly be said that welfare reform will have significantly reduced poverty in America.


Center for Civic Innovation.


CB 17 PDF (151 kb)


Welfare Reform Worked by June O’Neill, Wall Street Journal, 8-1-01
Surprise! The Welfare Reform Law Has Worked -- So Far,, 8-8-01
Reform law helped reduce welfare rolls The Washington Times, 8-7-01
Pride in Welfare Reform, The Post and Courier (Charleston, S.C.), 8-3-01
Another Look at Reasons for Positive Welfare Report, San Jose Mercury News, August 1, 2001
Welfare Reform’s Latest ‘A’ New York Post, 8-1-01
Reliance on Welfare Windfall May Prove Costly for State and City, Report Warns New York Times, 7-31-01
Study praises welfare reform Chicago Tribune, 7-26-01

Gaining Ground? is a comprehensive study of the changes in welfare and work participation among single mothers since the implementation of welfare reform. Contrary to the claims and expectations of reform critics, this study finds that welfare participation has dropped 50 percent and employment has risen by a roughly equivalent amount between August 1996, when the reform legislation was passed, and September 2000, with disadvantaged mothers showing the greatest gains. Moreover, it finds that provisions of the welfare reform law were responsible for more than 50 percent of these advances, while less than 20 percent were the result of the economic boom of the late 90’s.


About the Authors

Authors’ Acknowledgements

Executive Summary


A Brief History of Welfare Reform

Leaving Welfare, Going To Work: What The Data Show

Figure 1: Families Receiving AFDC/TANF, Years 1936–2000

Figure 2: Single Mothers Reduced Their Welfare Participation and Increased Their  Work Participation as Their Exposure to Welfare Reform Grew

A. Conceptual Issues

B. Data Sources

C. Single Mothers, Especially Disadvantaged Mothers, Have Left the Welfare Rolls  in Large Numbers

Age of Mother and Age of Youngest Child

Figure 3: The Decline in Welfare Participation is Largest for Single Mothers  in the Youngest Age Groups

Figure 4: The Decline in Welfare Participation is Largest for Single Mothers with Young Children

Schooling and Race

Figure 5: A Larger Welfare Decline is Observed for High School Dropouts

Figure 6: Welfare Participation Has Declined for All Race/Ethnicity Groups, but the Decline is Larger for Black and Hispanic Mothers

Marital Status

Figure 7: Welfare Participation Declined More for Never-Married Mothers Than for Previously Married Mothers

D. Single Mothers, Especially Disadvantaged Mothers, Greatly Increased  Their Work Participation

Marital Status

Figure 8: Single Mothers, Especially Never-Married Mothers, Experienced Greater Employment Gains Than Married Mothers

Age of Mother and Age of Youngest Child

Figure 9: Younger Single Mothers Made the Largest Employment Gains

Figure 10: Employment Increased the Most for Single Mothers with Young Children

Schooling and Race

Figure 11: Large Gains in Employment Made by the Least-Educated Mothers

Figure 12: Particularly Strong Employment Gains by Black and Hispanic Single  Mothers Have Narrowed Racial and Ethnic Employment Differentials

Summary on Changes in Welfare and Work

Welfare Reform is the Leading Cause of Recent Declines in Welfare  and Increases in Work

A. Research Strategy

B. Using Regression Analysis To Determine the Effect of TANF on Welfare  Participation and Work Participation

The Variables

Regression Results

Differences by Age Group

Table 1: Estimated Impact of Welfare Reform on the Percentage of Single Mothers Receiving Welfare and the Percentage Who Work

Differences by Education and Race

Table 2: Differences by Education in the Estimated Impact of Welfare Reform on  the Percentage of Single Mothers Receiving Welfare and the Percentage Who Work

Table 3: Differences by Race/Ethnicity in the Estimated Impact of Welfare Reform on  the Percentage of Single Mothers Receiving Welfare and the Percentage Who Work

C. Welfare Reform Is Responsible for Over Half of the Decline in Welfare  Participation Since 1996, and Over 60 Percent of the Rise in Work Participation

Table 4: The Contribution of Welfare Reform and the Decline in Unemployment to the Decline in Welfare Participation and the Increase in Work Participation of Single  Mothers by Age Group

Table 5: The Contribution of Welfare Reform and the Decline in Unemployment to the Decline in Welfare Participation and the Increase in Work Participation of Single  Mothers by Education

Table 6: The Contribution of Welfare Reform and the Decline in Unemployment to the Decline in Welfare Participation and the Increase in Work Participation of Single  Mothers by Race


Appendix A

Table A-1: AFDC/Waiver Policies Implemented in States

Table A-2: TANF Policies Implemented Between 1996 and 1998 by State

Appendix B: Data Issues: Caseload Data Versus the CPS

Figure B-1: Trends in the Number of Female Family Heads Receiving Welfare Are  Similar Under Different Definitions and Data Sources

Appendix C: Comparing the Characteristics of Welfare Recipients  and Other Women

Table C-1: Characteristics of All Women and Women with Children Under 18,  Ages 18–44, by Marital Status, Before and After Welfare Policy Changes

Table C-2: Characteristics of Single Mothers, Ages 18–44, by Their Welfare Status,  Before and After Welfare Policy Changes



Home | About MI | Scholars | Publications | Books | Links | Contact MI
City Journal | CAU | CCI | CEPE | CLP | CMP | CRD | ECNY
Thank you for visiting us.
To receive a General Information Packet, please email
and include your name and address in your e-mail message.
Copyright © 2015 Manhattan Institute for Policy Research, Inc. All rights reserved.
52 Vanderbilt Avenue, New York, N.Y. 10017
phone (212) 599-7000 / fax (212) 599-3494