TANF "Leavers" and Diversion Studies


TANF Leavers, Applicants, and Caseload Studies:

Preliminary Analysis of Racial Differences in Caseload Trends and Leaver Outcomes

By Elizabeth Lower-BaschOffice of the Assistant Secretary for Planning and Evaluation

Text Last Revised: December 2000

Main Page | Reports | Links | (FAQs) ]


Poverty Rates

Welfare Receipt

Findings from Welfare Leavers Studies


Directions for Further Research




The strong United States economy and changes in welfare and tax policy have combined to reduce welfare rolls to their lowest level in decades. However, there is some concern that minorities are not sharing equally in the benefits of these changes. This paper discusses trends in national data on both poverty rates and welfare receipt over the past 15 years. It then examines some preliminary evidence from state and national studies of families leaving welfare and uses them to generate some hypotheses to explain the trends.

[ Contents ]

Poverty Rates

A study of caseload trends benefits from first understanding the trends in poverty rates, since the welfare population is a subset of the overall poverty population. According to Census Bureau figures, the overall United States poverty rate fell from 12.7 percent in 1998 to 11.8 percent in 1999  a significant drop. This brings the poverty rate to its lowest point since 1979. The total size of the poverty population is now 32.3 million people, a decline from 34.5 million in 1998. These figures are based on the official poverty measure, which does not reflect the effects of the Earned Income Tax Credit (EITC). Including the EITC, the reduction in poverty would be even greater.

Minorities continue to have substantially higher poverty rates than Whites, but all racial groups are sharing in the positive trends. The poverty rate for Blacks was 23.6 percent in 1999, down from 26.1 percent in 1998, and the lowest rate since 1959. For Hispanics, the poverty rate was 22.8 percent in 1999, down from 25.6 percent in 1998.

Poverty rates for families with children are higher than poverty rates for all families   13.8 percent compared to 9.3 percent in 1999. Families with children that are headed by single women have even higher poverty rates  35.7 percent in 1999. As shown in Table 1, these patterns hold true across all racial groups. However, even among female-headed families, minorities are substantially more likely to be poor than whites. The good news is that, while the poverty rates for families with children remain unacceptably high, the progress since 1985 has been quite dramatic, particularly for the most disadvantaged families. Poverty rates for minority female-headed households have fallen by about 25 percent since 1985.

It is important to note that while the poverty rates for Black and Hispanic families with children have fallen significantly since 1985, the total number of poor Hispanic families has increased by almost 40 percent over that time period. This is because the total number of Hispanic families with children increased by almost 80 percent. So, even though a smaller fraction of Hispanic families with children were poor, because the base was so much higher, the total number of such families who were poor grew. Among Black families, the declining poverty rate and the increase in population offset each other, resulting in a small decrease in the number of poor Black families with children. The number of non-Hispanic White families with children remained approximately constant over this period, so the declining poverty rate resulted in a smaller number of poor non-Hispanic White families with children. As a result, Black and Hispanic families were an increasing fraction of all poor families, even as their poverty rates declined.

Table 1 Numbers of Families and Poverty Rates, United States, 1985  1999, by Race
 19851999Change 1985-1999
# Familes with children under 1824,9164,6362,97324,7845,5855,320-0.5%20.5%78.9%
# Poor familes with children under 182,7761,6709551,9841,6151,330-28.5%-3.3%39.3%
Poverty rate, familes with children under 1811.1%36.0%32.1%8.0%28.9%25.0%-27.9%-19.7%-22.1%
# Female headed families with children under 183,7372,2697714,2522,8921,35313.8%27.5%75.5%
# Poor female headed families with children under 181,2661,3364931,0791,333630-14.8%-0.2%27.8%
Poverty rate, female headed families with children under 1833.9%58.9%64.0%25.4%46.1%46.6%-25.1%-21.7%-27.2%
Figures in thousands. "White" means Non-Hispanic White.

Source: U.S. Bureau of the Census, Historical Poverty Tables, Table #4.

Finally, it is worth noting that the trends of the last 15 years are not driven by a change in the percentage of families with children that are headed by women. The fraction of all families with children that are female-headed increased slightly for Non-Hispanic Whites and for Blacks, and decreased slightly for Hispanics. Moreover, the poverty rate fell more for female-headed Hispanic families with children than it did for all Hispanic families with children.

[ Contents ]

Welfare Receipt

As a result of a number of factors, including welfare reform, the strong economy, and the EITC, welfare caseloads have declined dramatically since 1994. The national data reported to the Administration for Children and Families (ACF) by the states indicate that caseloads have declined for all racial groups, but there is some variation in the patterns. As shown in Figure 1, the number of Hispanic families on welfare almost doubled between 1985 and 1995, increasing from under 400,000 to over 1 million. During this period, the number of Black and White families on welfare also increased, although not as rapidly. The caseload peaked for these groups somewhat earlier as well   in 1993 for White families and in 1994 for Black families. Since 1995, the number of Hispanic families on welfare has declined somewhat, but has only returned to the levels of the early1990s. By contrast, the number of White and Black families on welfare has declined sharply, falling to levels not seen in decades.

Figure 1: Number of AFDC/TANF Families by Race, 1985-1999.

Source:  Department of Health & Human Services, Administration for Children and Families, Characteristics and Financial Circumstances of TANF Recipients, 1998, and earlier reports. Race is determined based on the race of the adult recipient; in child-only cases, race is determined based on the youngest child in the unit.

While the caseload trends for Black and White families are quite similar from 1985 through 1996, there appears to be some divergence in the post-welfare reform period 1996-1999, with the number of White recipient families falling faster (a 50.6 percent decline) than the number of Black recipient families (a 39.6 percent decline). It is too soon to tell whether this is the beginning of a trend, or simply a short-term variation.

As a result of these differing patterns of caseload increase and decline, there has been a significant change in the composition of the welfare caseload. As shown in Table 2, between 1985 and 1996, the major trend in the racial composition of the caseload was the increase in the proportion of the caseload that is Hispanic, which rose from 13.6 percent in 1985 to 20.7 percent in 1996. The proportion of the caseload that is Non-Hispanic White gradually declined from 40.8 percent in 1985 to 35.6 percent in 1995. The proportion of the caseload that is African-American declined from a high of 41.6 percent in 1985 to a low of 36.4 percent in 1994. This general trend has continued since implementation of welfare reform. In the most recent data, for FY 1999, Non-Hispanic White families made up 30.5 percent of the caseload, African-American families 38.3 percent, and Hispanic families, 24.5 percent.

As in the poverty statistics, it appears that the major factor behind the increasing proportion of Hispanic families in the welfare caseload is simply the large increase in the total number of Hispanic families in the U.S. population. This trend began before the implementation of welfare reform policies, and has continued since. Another factor which may have affected these trends is the Immigration Reform and Control Act (IRCA) of 1986, which granted legal status to many undocumented immigrants. By contrast, the provisions affecting immigrants in the 1996 welfare law were generally expected to have the effect of reducing participation in welfare programs by non-citizens and their family members.

Table 2 Distribution of AFDC/TANF Families by Race 1985  1999
Fiscal YearRace of Parent
1.  1997 data is for October 1996June 1997 only, due to the transition from AFDC to TANF Race is determined based on the race of the adult recipient; in child-only cases, race is determined based on the youngest child in the unit.

Source:  Department of Health & Human Services, Administration for Children and Families, Characteristics and Financial Circumstances of TANF Recipients, 1998, and earlier reports.

The overall national figures obscure a substantial amount of state-to-state variation. In particular, the Hispanic caseload is highly concentrated in a small number of states. In 1999, California accounted for an astonishing 45 percent of all Hispanic families receiving TANF, compared to 24 percent of all TANF families.  New York followed with 17 percent of Hispanic families (and 11 percent of all TANF families) and Texas had 9 percent of Hispanic families (and 4 percent of all TANF families). Thus, the story of what happened to Hispanic families receiving welfare is largely a story of what happened in these three states, and especially in California. In that light, it is worth noting that California has had substantially slower declines over the past few years in welfare caseloads than the national average across all three major racial groupings, possibly due to its later implementation of TANF. However, within California, caseloads have fallen much less among Hispanic families than among either Black or non-Hispanic White families, indicating that this is not a purely geographic story. (See Table 3.)

Table 3 Percentage Change in TANF Caseload, 1996  1999 by Race, for Selected States
United States-41.8%-50.6%-39.6%-31.5%
New York-32.1%-41.6%-17.3%-30.2%

When more data become available, additional work could be done to test these hypotheses and to examine the relative importance of different factors in explaining the shifts in caseloads. Some directions for further research are outlined in Section V of this paper.

[ Contents ]

Welfare Leaver Studies

A number of studies have begun to examine the population exiting welfare (known as welfare "leavers") and what happens to them after their exit, using administrative records from both the welfare and the employment insurance system as well as surveys. A few of these studies examine differences by racial group, including studies in Arizona, Cuyahoga county (Cleveland, Ohio), Florida, Illinois, Missouri, Washington state, Wisconsin and the Urban Institute's National Survey of America's Families (NSAF). While each of these studies answers slightly different questions, by looking across studies it is possible to piece together a larger picture. However, it is important to note that almost none of the studies reported tests of statistical significance for the differences between racial and ethnic groups. Therefore, these findings must be treated with caution.

In addition, the findings from the recent leaver studies are compared to those from a few studies which examined racial differences in outcomes for welfare recipients and leavers during the pre-TANF period. These findings are important in helping us understanding whether these patterns have changed since welfare reform.

Is there a difference in rates of exit by race/ethnicity?

The studies that have compared welfare leavers to all TANF cases, or to cases that did not close, have generally found that African-Americans were underrepresented among leavers. In other words, they were less likely to leave welfare than Non-Hispanic Whites. In Washington State, African-Americans were 14 percent of those continuously on TANF from November 1998 to April 1999 and just 8 percent of those leaving TANF in October 1998. In Missouri, Blacks were 35.0 percent of leavers during the fourth quarter of 1996, compared to 46.0 percent of all welfare cases. Illinois compared leavers from the fourth quarter of 1997 to cases that were open in July 1997 and remained open for the next 18 months. African-Americans constituted 49.5 percent of the leavers, as compared to 72.0 percent of the cases that remained open for 18 months. In Wisconsin, from July 1995 to July 1996, 60.8 percent of white recipients left welfare as compared to 36.3 percent of African-American recipients. This study found that even after a range of other characteristics were controlled for, African-Americans were still significantly less likely to leave welfare than Non-Hispanic Whites.

However, the one national source of data, the Urban Institute's National Survey of America's Families, or NSAF, indicates that Non-White Non-Hispanics were only slightly underrepresented among leavers compared to current recipients. Non-White Non-Hispanics constituted 34.7 percent of those who had received welfare benefits at some point since 1995, but were not on assistance when interviewed in 1997, and 36.6 percent of those who were on welfare at the time of the interview. Similarly, in Arizona, in the first quarter of 1998, Blacks were 9 percent of open cases and 10 percent of closed cases.

The pattern for Hispanics differs from study to study. According to the NSAF, Hispanics were substantially underrepresented among welfare leavers (13.1 percent) compared to current recipients (21.7 percent in 1997). Several state studies found Hispanics about evenly represented among current recipients and leavers. In Arizona, in the first quarter of 1998, Hispanics accounted for 33 percent of open cases and 35 percent of the leavers. In Illinois, Hispanics were 9.4 percent of the leavers and 8.8 percent of the 18-month stayers. In Wisconsin, 45.7 percent of Hispanic recipients left welfare, placing them in the middle compared to African-Americans and non-Hispanic Whites. The only leavers study so far finding Hispanics overrepresented among leavers is from Washington State, where Hispanics were 13 percent of those leaving TANF in October 1998, versus 7 percent of those remaining of TANF from November 1998 to April 1999.

This complexity of findings may reflect with the fact that the Hispanic population in the United States is both highly diverse and unevenly distributed. As a result, they may be affected differently by the changing policy environment, e.g., Hispanic groups whose members are mostly citizens will not be affected by the changes in policies for new immigrants. Moreover, the findings are not all from the same time period; it is possible that the policy changes affecting immigrants caused a short-term shock to the welfare caseloads which would only have been captured by some of the studies.

Because relatively little attention was paid to welfare leavers in the past, there are no directly comparably figures from before welfare reform, so it is difficult to know whether these findings reflect a change from the pre-PRWORA period. However, it is known from the extensive literature about welfare dynamics that African-Americans were disproportionately represented among long-term welfare recipients. This would suggest that even before the implementation of welfare reform, African-Americans might have been underrepresented among welfare leavers.

Exits from TANF may be an unalloyed good, when they are due to increases in income which allow a family to achieve self-sufficiency, or may be a matter of concern, particularly when they are due to sanctions for non-compliance and when the family does not appear to have a replacement source of income. One disturbing finding is that those studies which examined reason for exit found that minorities were generally more likely than Whites to have their cases closed due to sanctions rather than earnings (according to administrative data). In Arizona, African-Americans were 13 percent of cases closed due to sanctions, and just 9 percent of cases closed due to earnings. Hispanics were 39 percent of cases closed due to sanctions and 33 percent of cases closed due to earnings. Interestingly Native Americans were just 8 percent of cases closed due to sanctions and 15 percent of cases closed due to earnings. The report notes that Native Americans are more likely to live in extremely high unemployment areas (greater than 50 percent unemployment) and therefore to be exempted from the job search requirements, which sometimes lead to sanctions. Similarly in Illinois, cases closed due to non-cooperation are more likely to be minority (61.6 percent African-American, 10.6 percent Hispanic, and 26.8 percent Non-Hispanic White) compared to cases closed due to increased income (52.4 percent African-American, 8.7 percent Hispanic, and 28.3 percent Non-Hispanic White).

Interestingly, when asked why their cases were closed, the majority of Arizona respondents of all races said that they got a job or a better job (63.8% of African-Americans, 50.0% of Native Americans, 55.7% of Non-Hispanic Whites, and 52.9% of Hispanics). African-Americans were actually somewhat less likely than Whites to report that their cases were closed due to non-compliance (7.6 percent of African-Americans v. 9 percent of Non-Hispanic Whites). Similarly, the Washington study also found little difference between Whites and non-Whites in their self-reported most important reason for leaving welfare. For both groups, two-thirds of respondents said that increased income through employment was their primary reason for leaving welfare, far ahead of the second answer, increased income through other sources (selected by only 8-9 percent of respondents).

This gap between the administrative reasons for closing and the reasons reported by the respondents suggests that (at least in Arizona) African-Americans who got jobs may have been less likely to report their income to the TANF agency and more likely to ignore requests for paperwork. If so, this may reduce their receipt of other services, such as transitional Medicaid, for which they are eligible.

Are there racial/ethnic differences in outcomes after leaving welfare?

A few of the studies of welfare leavers have gone further, and examined racial/ethnic differences in outcomes among the population leaving welfare. When studies reported the same outcomes, these are shown in the tables below. In general, these are simple cross-tabulations of outcomes (employment, earnings and welfare recidivism) by racial groupings. These do not account for other factors which may correlate with race, and which may also affect outcomes, such as geographic location, marital status, household composition, or education. The possible role of these other factors in explaining the differences in outcomes are discussed in more detail in Part IV. When studies reported different but related measures, they are discussed in the text.

The studies that examined employment and earnings of welfare leavers using administrative data generally found that African-Americans leaving welfare were more likely than Non-Hispanic Whites to be employed one quarter after exit (see Table 4).

Table 4 Employment Rates One Quarter After Exit from Welfare, by Race
 Georgia 1997Arizona 1996Q4Arizona 1998Q1Missouri 1996Q4Cuyahoga 1996Q3Illinois July 97  Dec 98

Similarly, Florida found that the average quarterly employment rate was 59.7 percent for Blacks, compared to 45.9 percent for Whites. Blacks were also more likely to be ever employed during the first two years after exit and more likely to be employed in all four quarters after exit. Illinois was an exception to this pattern, in that employment rates were slightly higher for Whites than for African-Americans. This difference disappeared entirely when other demographic characteristics were included in a regression predicting the likelihood of wage earnings in the quarter after exit.

African-American leavers also generally had higher median earnings (see Table 5). For example, in Cuyahoga, median earnings among those employed was $2,772 for African-Americans versus $2,112 for Whites. In Florida, the average annual earnings were $7,037 for Blacks compared to $6,351 for Whites.

Table 5Median Earnings One Quarter After Exit from Welfare, by Race
 Arizona 1996Q4Arizona 1998Q1Cuyahoga 1996Q3Missouri 1996Q4Wisconsin July 95  July 96

Hispanics' relative status varied from study to study. In Arizona, Hispanics, like African-Americans were more likely to be employed than Non-Hispanic Whites, while in Cuyahoga, Hispanics were even less likely to be employed than Non-Hispanic Whites. In the Illinois regression, being Hispanic was associated with a slightly higher likelihood of having earnings in the quarter after exit, holding a variety of other factors constant. The Wisconsin study and one Arizona study found median incomes for Hispanics between those for Non-Hispanic Whites and those for African-Americans, while the Cuyahoga study found that Hispanics had the highest median incomes of the three racial groups, and another Arizona study, using the same methodology for a later cohort, found Hispanic median incomes below those of either African-Americans or Non-Hispanic Whites. In Florida, Hispanics had employment rates about equal to Non-Hispanic Whites, but average earnings of $7,732  greater than either African-Americans or Non-Hispanic Whites.

Yet, while African-American welfare leavers are more likely to be employed and have higher median earnings, these studies also show that African-Americans are also more likely to return to welfare within one year (see Table 6).  In Wisconsin, statistical analysis confirmed that even after other characteristics were controlled for, minorities were significantly more likely to return to welfare. Illinois reported on recidivism somewhat differently, but also found that African-Americans constituted 68.7 percent of cases that returned to welfare within 6 months of exit, but just 50.8 percent of those that remained off for six months. In a regression model, being African-American was associated with a 79 percent greater chance of returning to TANF within 6 months, compared to Whites.

Table 6 Percentage Returning to Welfare within One Year of Exit from Welfare, by Race
 Arizona 1996Q4Cuyahoga 1996Q3Missouri 1996Q4Florida 1997Q2Wisconsin July 95  July 96

Across these studies, the return rates for Hispanics generally fell between those of African-Americans and those of Non-Hispanic Whites, although in Florida the Hispanic recidivism rate was essentially the same as that for Non-Hispanic Whites. In Maryland, Hispanics constituted 78.5 percent of those who had returned to welfare at the three-month follow up point, as compared to 69.7 percent of those who left and remained off. However, in Illinois, Hispanics were underrepresented among those who returned to TANF (7.8 percent of returners, v. 9.9 percent of those who remained off for six months) and this finding was confirmed by the regression, which found being Hispanic associated with a reduced likelihood of returning to TANF within 6 months, holding a variety of other factors constant.

These findings are consistent with pre-TANF patterns analyzed by Edin and Harris (1999). Looking at data from the PSID, they found that, of those welfare recipients who left welfare for work between 1983 and 1988, 29 percent of the African-Americans but just 12 percent of the Whites returned to welfare within one year. Over the full five-year period studied, 65 percent of African-Americans and 44 percent of Whites returned to welfare. This was true even though the African-Americans who left welfare for work had more education, more pre-exit work experience, were less likely to have a pre-school age child, had similar earnings, and were more likely to work continuously since exit.

[ Contents ]


These simple cross-tabulations reveal patterns of racial difference but, without further analysis, they provide no explanations for why these differences are occurring. One question of particular interest is why African-Americans leaving welfare generally have higher employment rates and earnings than Whites, but also higher rates of return to welfare Answering this question will require further analysis, including more detailed examinations of the ways in which race and ethnicity may be picking up the effects of such other characteristics as education, marital status, English-language abilities, urban or rural location, or other reasons for exit. Since the population exiting welfare is highly diverse, with some leaving voluntarily due to increased earnings or other means of support and others being forced off by sanctions, there is no reason to assume that those who are employed at exit are the same individuals who are returning to welfare. In the absence of more detailed analyses, we are only able to speculate at some possible explanations.

There is a long and varied literature on reasons why minorities are disproportionately likely to be poor and to receive welfare. Among the most commonly cited reasons are discrimination, lack of human capital (either education or work experience), family structure and social networks, and isolation in central cities. In this section, each of these reasons are considered in turn.


It is important to mention that one reason why different populations may have different outcomes is that they may have experienced different treatments. Under welfare reform, many states have moved away from centralized bureaucratic systems, which rigidly treated all recipients the same, to more flexible systems with substantial opportunity for caseworker discretion. This increases the possibility for differential treatment based on race or ethnicity. For example, one study in Virginia, with an admittedly limited sample size and methodological issues, found that Black welfare recipients reported receiving less discretionary transportation assistance and less caseworker support than White recipients.  In addition, the new eligibility rules based on citizenship status may be disproportionately affecting Hispanics and Blacks, particularly if caseworkers, overwhelmed by the complexity of these rules, are using race or ethnicity as a proxy for immigration status.

While this question is not directly addressed in the leavers studies analyzed here, some of the findings reported suggest that minorities may be particularly vulnerable to communications failures between recipients and caseworkers. For example, in the Arizona study, Hispanics and Native Americans were the only groups to report "did not complete interview process or provide documentation" as one of the top three reasons for case closures, and Blacks were the only group to report "avoiding hassles" in the top three reasons. Blacks were somewhat less likely than Whites to report "noncompliance with TANF requirements" as the reason their cases were closed, even though they were more likely to be listed in administrative records as having their cases closed due to sanctions. While far from conclusive, this suggests that the Blacks recipients may have had worse relationships with their caseworkers or the welfare agency.

Race may also affect the treatment received by recipients in a more indirect way, by affecting the policies chosen by policymakers at the state and local level. To the extent that welfare recipients are perceived as predominantly of a different racial or ethnic group than the majority of the taxpaying/voting citizens of a state, lawmakers may be more ready to impose stringent conditions on welfare receipt. In particular, an analysis by Sanford Schram of Bryn Mawr suggests that states are more likely to adopt full-family sanctions, time limits, and family cap policies the greater the percentage of African-Americans among their welfare caseloads. However, this does not account for the differences in outcomes by race within a jurisdiction. Moreover, the observed correlation would tend to cause greater caseload declines among minorities than among whites, while the reverse has actually occurred. Thus, there are clearly other factors involved.

Human capital

It is particularly interesting to learn that African-American leavers are more likely to be employed than White leavers, and have higher average earnings given that education is a strong predictor of both employment and earnings, and that African-Americans and Hispanic welfare recipients are more likely to lack a high school degree than Non-Hispanic White recipients. In 1994, just 33 percent of Non-Hispanic White recipients lacked a high school degree, while 40 percent of Black recipients and 64 percent of Hispanic recipients lacked one.

In general, the leavers studies have not looked at educational levels, but one that did, in Washington state, found that among leavers, Whites had an average of 11.8 years of education, to Non-Whites' 11.2 years. By contrast, in their pre-TANF work based on a national sample, Edin and Harris found that of those who left welfare for work, 49 percent of Whites but only 23 percent of Blacks had less than a high school diploma.

Family structure and social networks

Marital status, child support and earnings of other family members

There are a number of ways in which family structure and social networks could influence both the probability of leaving TANF and outcomes for those families who leave welfare. One hypothesis that is consistent with the patterns described above is that White recipients are more likely to have sources of support other than their own earnings, such as earnings of a household member or child support payments, that enable them to leave welfare. These sources of support may be more stable than the jobs that welfare recipients are able to achieve, which are often temporary or offer variable hours of work.

This hypothesis is partially supported by evidence from the Washington state study, which uses survey data to provide a richer picture of the household situation of the former recipients. Like the studies cited above, this study found that non-Whites had somewhat stronger employment outcomes than Whites. However, more non-Whites saw themselves as very likely to return to TANF within 6 months (12 percent v. 6 percent). Almost twice as many White respondents expected to receive child-support as non-Whites and, on average, they received a higher share of the expected amount. Whites were also twice as likely (31 percent v. 15 percent) to live with another worker with earnings in the past month, and the amount earned by other workers was almost three times as much (mean of $460 v. $164). Overall, mean family income for the White families was 14 percent higher than for non-Whites ($1,444 v. $1,235).

In Florida, however, White leavers were less likely to report receiving child support (30.4 percent) than either Black (37.0 percent) or Hispanic (38.6 percent) leavers. There were not strong racial patterns in the likelihood of receiving help with bills from someone living with the respondent (27.3 percent for Blacks, 31.2 percent for Hispanics, and 33.1 percent for Whites) or from someone not living with the respondent (33.5 percent for Blacks, 35.4 percent for Hispanics and 32.2 percent for Whites).

This hypothesis is, however, consistent with pre-TANF findings. Looking at the 1983-1988 PSID sample, one-third of white women who left welfare for work subsequently married, and over one-fifth cohabited. By contrast, less than one in ten African-American women either married or cohabited after exit. This may help explain why 80 percent of the African-American households remained poor in the year that the mothers left welfare for a job, compared with about 50 percent of the white households, even though the earnings for the mothers were about equal. These data also suggest that marriage (or cohabitation) and work are not alternative strategies, but rather complementary strategies: those White mothers with the best educational credentials and highest earnings were also the most likely to marry or cohabit.

Looking at a different pre-TANF sample, the control group for the Riverside, CA site in the National Evaluation of Welfare to Work Strategies, a similar pattern is evident. Blacks were more likely than Whites to be ever employed in the first two years after random assignment (51.7 percent v. 43.7 percent), but had approximately the same earnings over this time period. Yet, whites had higher overall household income than Blacks in the month before the two year interview ($1,488 v. $1,284), because they received a greater share of their total household income from other household members (31.0 percent v. 22.8 percent). Whites were twice as likely as Blacks to be married (15.6 percent v. 7.5 percent), cohabiting (15.0 percent v. 5.9 percent), or receiving child support (13.5 percent v. 6.4 percent).

Family size

Another possible explanation is that, on average, African-American and Hispanic families receiving welfare have slightly larger families than white families receiving welfare, which means that they require higher income levels in order to support themselves off welfare, and thus can not exit welfare until they have achieved those levels. Thus, for example, in FY 1998, white families constituted 32.7 percent of families receiving welfare, but white children were only 28.5 percent of children on welfare. With more children there is also a greater probability of a child's needs interfering with work, such as a sick child needing care. In the Florida study, 38.5 percent of Hispanic leavers and 35.1 percent of Black leavers, compared to 31.1 percent of White leavers reported that they missed a day of work in the past month because of a child care problem.

A few of the leavers studies examined outcomes by family size. The Cuyahoga study found that median earnings in the first quarter after exit were somewhat higher for those with three children ($2,745) than those with two children ($2,688), which were in turn higher than those with just one child ($2,553). However, those with one or two children had higher employment rates (61.1 percent and 62.0 percent) than those with three children (53.5 percent).

The likelihood that family size contributes to recidivism is supported by the Maryland study, which found that those who returned to welfare within three months had an average of 1.96 children (median of 2) while those who stayed off welfare for at least three months had an average of 1.76 children (median of 1). This difference was statistically significant at the .001 probability level. The Cuyahoga study found a less consistent pattern, with families with one child having the lowest recidivism rate at one year (34.6 percent), followed by families with three children (37.4 percent) and then families with two children (40.0 percent).

Child-only cases

A third way in which family structure may affect the proportion of the welfare caseload that is non-white may have increased in some states is the presence of a large number of child-only cases, which are more likely to be non-white than other TANF cases. In FY 1998, 28.2 percent of child-only cases were white, 39.9 percent were Black, 26.7 percent were Hispanic, 2.3 percent were Asian, and 1.4 percent were Native American. This disparity may be because Black children are more likely than children in other racial groups to be cared for by relatives other than their parents (such cases are 39 percent of child-only cases) and because Hispanic children are more likely to have parents who are not eligible for welfare on account of their immigrant status (16 percent of child-only cases).

Child-only cases are less likely to leave welfare for work than parent-present households, and have therefore declined much more slowly than the overall caseload in recent years. As a result of this trend, even if the racial composition of other cases held constant, the overall racial composition of a state with a large child-only population would probably become more non-white. As this hypothesis suggests, those states with the highest proportion of child-only cases are more likely to have experienced large drops in the fraction of the caseload that is white in the past few years.

Isolation in central cities

One likely explanation is that minority welfare recipients are disproportionately likely to live in central cities, where welfare caseloads have not declined as rapidly as in other areas. A Brookings Institution survey of the 89 urban counties that contain the 100 largest U.S. cities found that the urban counties' welfare caseloads dropped by 40.6 percent from 1994 to 1999, compared to a national caseload decline of 51.5 percent.

Such cities often have higher rates of unemployment than the surrounding areas, and inner city residents may find it difficult to get to the areas where jobs are available. Thus, for example, in Florida, 46.9 percent of Black leavers and 44.8 percent of Hispanic leavers, but only 32.6 percent of White leavers indicated that they could have had a better job if they had a better means of transportation. Urban welfare offices also typically have larger caseloads, which may make it more difficult to accomplish the culture change that is a major component of welfare reform. In some cases, states began implementing welfare reform later in urban areas than in other parts of their state.

A few of the leavers studies examine the interaction between race and urban location; the general pattern is that the effect of race is greatly diminished, although not entirely eliminated, when urban location is also considered. For example, the Illinois report notes that the finding that Whites were more likely to be employed after exit than Blacks or Hispanics totally disappeared when the leavers were divided by Cook County (Chicago) vs. Downstate.

Table 7 Percent Employed During First Quarter After Exit Illinois Leavers: July 1997-December 1998
Cook County48.451.951.5

While the other studies do not do this detailed an analysis, the Missouri report notes that the St. Louis area (both city and county) accounts for just 24.5 percent of leavers, and 35.9 percent of all recipients, probably due to its higher unemployment rate. Only 11 percent of White recipients live in the St. Louis area, while 64 percent of Black recipients do.

Within urban areas, minorities, especially Blacks, are disproportionately likely to reside in areas of concentrated poverty and welfare receipt. This can be shown by examining the racial composition of the four counties being studied as part of the Project on Devolution and Urban Change. (See Table 8.) As in many large urban counties, Blacks and Hispanics constitute a large majority of the caseload in these counties, but Whites still constitute between 48.4 percent (in Cuyahoga) and 4.3 percent (in Miami-Dade) of the caseload. However, when the sample is limited to welfare recipients in high poverty (>40%) or high welfare (>20%) census tracts, the proportion of White recipients falls dramatically.

Table 8 Racial Composition of Welfare Caseload in Four Urban Counties Overall (unshaded) and High Poverty/High Welfare Census Tracts (shaded)
Cuyahoga County, OHLos Angeles County, CAMiami-Dade County, FLPhiladelphia County, PA
Black, Non-Hispanic47.0%79.8%24.8%52.4%53.0%65.1%68.3%74.8%
White, Non-Hispanic48.4%11.7%14.3%2.1%4.3%2.2%12.0%5.2%
Data from: Quint et al., Big Cities and Welfare Reform: Early Implementation and Ethnographic Findings from the Project on Devolution and Urban Change (New York, NY: MDRC, 1999), Table 2.2 and unpublished table of "Selected Background Characteristics of the Urban Change 1998-1999 Survey Sample, by Site." Survey sample was drawn to be representative of May 1995 welfare recipients residing in high poverty (>40%) or high welfare (>20%) census tracts.

[ Contents ]

Directions for Further Research

The findings summarized in this paper provide more questions than answers. The following are offered as some possible directions for further research, but are not intended to be a comprehensive list.

  1. Examination of the effects of increased exits from welfare, reduced entries to welfare, and returns to welfare on the racial composition of the caseload. Is the evidence from the leaver studies consistent with the story told from the caseload trends?
  2. Reanalysis of the caseload trends, looking only at cases including adults. The dynamics of child-only cases appear to be sufficiently distinct that they may obscure trends in the one- and two-parent caseloads.
  3. Analysis of administrative data for more states. Are the patterns consistent across states? Are there regional patterns which may be affecting racial/ethnic groups differently?
  4. Analysis of other groups, such as Asians and Native Americans, particularly in those states where they constitute a higher proportion of the welfare caseload. Of the studies examined in this report, only the Arizona studies reported information for these subgroups.
  5. More sophisticated analyses of the effects of race and other characteristics, such as reason for exit and age and number of children, on outcomes, using regression analysis or more detailed cross-tabulations.
  6. Analysis of survey data, particularly data on household composition and sources of household income. Is it generally true that White leavers are more likely to have additional sources of income? What impact does this have on family and child well-being?
  7. Analysis of pre-TANF welfare exit and return patterns. Are the patterns found here a new phenomenon under TANF, or were minorities leaving welfare always more likely to return?
  8. More detailed analysis of caseload trends among Hispanics and Asians to understand the effects of immigration-based changes in eligibility on participation among these groups.

Comments on this paper or responses to any of the questions posed would be welcomed, and should be addressed to Elizabeth Lower-Basch at (202) 690-6808 or Elizabeth.Lower-Basch@hhs.gov

[ Contents ]


Allen, Katherine and Kirby, Maria. Unfinished Business: Why Cities Matter to Welfare Reform. Brookings Institution, Center on Urban and Metropolitan Policy, July 2000.

Cancian, Maria, Haveman, Robert, Kaplan, Thomas and Wolfe, Barbara. Post-Exit Earning and Benefit Receipt among Those Who Left AFDC in Wisconsin. Institute for Research on Poverty, University of Wisconsin-Madison, October 30, 1998.

Coulton, Claudia and Verma, Nandita. Employment and Return to Public Assistance Among Single, Female Headed Families Leaving AFDC in Third Quarter, 1996, Cuyahoga County, Ohio. Cuyahoga Work and Training, May 1999.

Crew, Robert E, Jr., Eyerman, Joe, Graham, Justin and McMillan, Nancy. Tracking the Outcomes of Welfare Reform in Florida for Three Groups of People. Florida State University, October 2000.

Du, Jean with Debra Fogarty, Devin Hopps and James Hu. A Study of Washington State TANF Leavers and TANF Recipients: Findings from the April-June 1999 Telephone Survey (Final Report). Department of Social and Health Services, Economic Services Administration, Office of Planning and Research, February 23, 2000.

Edin, Kathryn and Harris, Kathleen Mullan. "Getting Off and Staying Off: Racial Differences in the Work Route Off Welfare." In Latinas and African American Women at Work, edited by Irene Browne. New York: Russell Sage Foundation, 1999.

Foster, E. Michael. Amended Quarterly Progress Report: Outcomes for Single Parent Leavers by Cohort Quarter. Georgia Department of Human Resources, Evaluation and Reporting Section, Jan-March 1999.

Genser, Jenny. Who is Leaving the Food Stamp Program? An Analysis of Caseload Changes from 1994 to 1997. U.S. Department of Agriculture, Food and Nutrition Service, Office of Analysis, Nutrition and Evaluation, March 1999.

Gooden, Susan T. "All Things Not Being Equal: Differences in Caseworker Support Towards Black and White Welfare Clients." Harvard Journal of African American Public Policy. Volume IV, 1998, pp. 23-33.

Harknett, Kristen. Working and Leaving Welfare: Does Ethnicity Matter? Unpublished paper presented to Population Association of American conference, March 25, 2000.

Julnes, G., Halter, A., Anderson, S., Frost-Kumpf, L., Schuldt, R., Staskon, F., and Ferrara, B. Illinois Study of Former TANF Clients, Final Report. Institute for Public Affairs, University of Illinois at Springfield and School of Social Work, University of Illinois at Urbana-Champaign. July 2000.

Loprest, Pamela. Families Who Left Welfare: Who Are They and How Are They Doing? (ANF Discussion Paper 99-02) The Urban Institute, 1999.

MaCurdy, Thomas, Mancuso, David, and O'Brien-Strain, Margaret. The Rise and Fall of California's Welfare Caseload: Types and Regions, 1980-1999. San Francisco, CA: Public Policy Institute of California, 2000.

Ryan, Sharon., Choi, Seung-Whan, Theilbar, Melinda. Zhang, Wei and Park, Do-Yeun. Draft Final Report: Phase I (1996 Cohort) University of Missouri-Columbia. Forthcoming.

Savner, Steve. "Welfare Reform and Racial/Ethnic Minorities: The Questions to Ask." Poverty and Race. Poverty and Race Research Action Council, Volume 9, Number 4, Page 3. (July/August 2000).

Welfare and Child Support Research and Training Group, School of Social Work, University of Maryland-Baltimore. Life After Welfare: Fourth Interim Report. Family Investment Administration, Maryland Department of Human Resources, October 1999.

Westra, Karen L. and Routley, John. Cash Assistance Exit Study: First Quarter 1998 Cohort. Arizona Department of Economic Security, Office of Research and Evaluation, May 5, 1999.

Westra, Karen L. and Routley, John. Arizona Cash Assistance Exit Study: Cases Exiting Fourth Quarter 1996. Arizona Department of Economic Security, Office of Research and Evaluation, July 1999.

Westra, Karen L. and Routley, John. Arizona Cash Assistance Exit Study: First Quarter 1998 Cohort, Final Report (Administrative and Survey Data Results). Arizona Department of Economic Security, Office of Evaluation, January 2000.

Racial Differences in TANF Exits and Life after Welfare: Results from Washington's TANF Exiter Survey. Presented at August 1999 National Association for Welfare Research and Statistics conference.


1.  The fraction of the national caseload that is made up of Asian and Native American recipients is too small to allow for detailed analysis. However, there is a great deal of variation among states. Native American families make up more than 50 percent of the TANF caseload in two states (North Dakota and South Dakota), and in another two states they make up 25-49 percent (Alaska and Montana). Asians make up 51.7 percent of the caseload in Hawaii, but are not more than 10 percent of the caseload in any other state. [Return to text.]

2.  For a detailed explanation of these provisions, see: Summary of Immigrant Eligibility Restrictions under Current Law as of 4/15/99, at:http://aspe.hhs.gov/hsp/immigration/restrictions-sum.htm  [Return to text]

3.  As of March 2000, California contained 12 percent of all households, and 29 percent of all households headed by Hispanics. (March 2000 Current Population Survey data, as tabulated by Rebecca L. Clark, NICHD). California has a disproportionate share of the national TANF caseload both because it has a somewhat higher than average poverty rate (15.3 percent during the three year period 1997-1999, compared to a national average of 12.6 percent) and because it has among the most generous TANF income standards for determining eligibility.  [Return to text.]

4.  Susan T. Gooden, "All Things Not Being Equal: Differences in Caseworker Support Towards Black and White Welfare Clients," Harvard Journal of African American Public Policy. (Volume IV, 1998, pp. 23-33.)  [Return to text.]

5.  Emergency TANF Data Report files.  [Return to text.]