The National Evaluation of the Welfare-to-Work Grants Program, Final Report

Chapter V:
The Well-Being of WtW Enrollees

[ Main page of Report | Contents of Report ]

Contents

1. Receipt of Assistance
   1. Assistance from Private Sources
   2. Assistance from Public Sources
   3. Independence from Tanf Through Employment
2. Household Income and Poverty
   1. Household Income
   2. Poverty Rates
   3. Material Distress
3. Hardships During the Second Year
   1. Homelessness
   2. Lack of Health Insurance

• Endnotes

The premise underlying the WtW grants program is that hard-to-employ individuals will be less dependent on TANF and their well-being will improve if they obtain jobs. The preceding chapter presented findings that two-thirds or more of WtW enrollees across the study sites obtained jobs sometime during the year following program entry. However, their employment rates were lower during the second year. By the end of that year, fewer than half of enrollees were employed in 8 of the 11 sites. Those who were employed tended to hold jobs that offered slightly higher compensation in the form of wages and fringe benefits than was true the year before. The net result of these trends was that the mean earnings of enrollees two years after program entry were about the same as one year after entry.

WtW enrollees did become less dependent on TANF following program entry; but for most, incomes did not improve to levels that allowed them to escape poverty. As described in Section A of this chapter, rates of TANF receipt by WtW enrollees fell dramatically during the two years after program entry, but enrollees remained highly dependent on assistance from a wide range of government programs. Despite earnings from employment, Section B shows that two years after program entry, two of every three enrollees had household incomes below the poverty threshold. These high poverty rates were not, however, accompanied by high levels of self-reported material distress. Section C reports on two specific hardships — homelessness and lack of health insurance — experienced by some enrollees. Rates of homelessness were high in several study sites but fell over time. On the other hand, the proportion of enrollees without health insurance coverage increased in most sites as time passed.

A. RECEIPT OF ASSISTANCE

WtW enrollees and their families were highly dependent on assistance from private and public sources two years after program entry; however, their dependency did lessen over the post-entry period. This section reports findings on the receipt of assistance based on data obtained from the study’s 12- and 24-month follow-up surveys and from state administrative records.

1. Assistance from Private Sources

WtW enrollees relied heavily on help from nongovernmental sources of support. They were especially likely to receive assistance from family and friends. During the second year after program entry, about two of every three enrollees in each of the study sites received gifts or loans of money, help with transportation, or other types of support from extended family members and friends (Exhibit V.1). They were much less likely to receive assistance from community organizations; only about one in every three or four enrollees received assistance from food pantries, crisis centers, or other organizations that provide goods and services to needy individuals and families.

The trend over time in the receipt of private-source assistance was downward. Across the 11 study sites, rates of receipt of assistance from family and friends or from community organizations were roughly 7 percentage points lower during the second year after program entry than the first (Appendix Exhibit B.10).⁽⁵⁰⁾ This may reflect movement by some enrollees beyond the disruptive events that precipitated their entry into WtW at the beginning of the previous year.

2. Assistance from Public Sources

Two years after entering WtW, most enrollees were still receiving some form of publicly funded assistance (Exhibit V.2). Only in Baltimore County and Milwaukee were rates of receipt of such assistance lower than 70 percent.⁽⁵¹⁾ By wide margins, food stamps were the most common form of government assistance; they were received by more than 60 percent of enrollees in all but the two aforementioned sites (Appendix Exhibit B.11). TANF and SSI/SSDI were the second or third most common forms of government assistance in all of the study sites except Milwaukee, where the noncustodial parents who had enrolled in the NOW program were more likely to receive Social Security retirement or survivors’ benefits and UI benefits than TANF.

TANF and Food Stamps. This study gathered and analyzed data on TANF and food stamp benefits from administrative files maintained by the states in which the study sites were located. These data provide a more complete picture than do survey data of changes over time in the receipt of benefits from these two programs. For each site, Exhibit V.3 presents quarterly rates of TANF and food stamp receipt calculated for the period beginning four quarters prior to program entry and ending eight quarters after program entry. Our discussion of this exhibit focuses on the interval from the quarter of program entry to the eighth quarter after entry and excludes the Milwaukee site due to its unique target population.⁽⁵²⁾

Receipt of TANF by WtW enrollees fell sharply during the two years following program entry. In the ten study sites that are the subject of this discussion, rates of TANF receipt by WtW enrollees during the quarter of program entry ranged from a low of 38 percent in Baltimore County to a high of 100 percent in West Virginia (Exhibit V.3). Eight quarters later, the distribution of recipiency rates had shifted markedly downward, ranging from 10 percent to 64 percent. The relative reductions in rates of TANF receipt were very large in most of the study sites.⁽⁵³⁾ In seven sites — Baltimore County, Boston, Chicago,⁽⁵⁴⁾ Ft. Worth, St. Lucie County, West Virginia, and Yakima — enrollees were at least 50 percent less likely to receive TANF eight quarters after program entry than in the quarter of entry. In the remaining three sites — Nashville, Philadelphia, and Phoenix — the reductions ranged from 30 to 40 percent.

This was a period when families in general were leaving TANF, so the reductions in recipiency by WtW enrollees were not unusual. Administrative data for all families receiving TANF in nine of the study sites during a selected quarter in the WtW enrollment period show relative reductions in recipiency rates over the ensuing eight quarters in excess of 40 percent (Appendix Exhibit D.2.c).⁽⁵⁵⁾ Therefore, the reductions in TANF receipt by WtW enrollees in the study sites should not be attributed to the local WtW programs; many of the enrollees in those programs would probably have left TANF even if they had not entered WtW.

Receipt of food stamps was relatively stable as TANF recipiency plummeted following WtW entry. The reduction in the rate of receipt of food stamps between Quarter 0 and Quarter 8 averaged about 30 percent across the 10 study sites and was generally much smaller than the reduction in the receipt of TANF (Exhibit V.3).⁽⁵⁶⁾ The Yakima site was typical; there, the rate of TANF receipt fell by 54 percent during the two years following WtW entry, whereas the rate of receipt of food stamps fell by only 27 percent. The West Virginia site, where TANF receipt fell by 68 percent while food stamp receipt fell by only 18 percent, provides the sharpest example of the relative stability of food stamp receipt. Only in Boston were the relative reductions in the rates of receipt of these two forms of government assistance following WtW entry roughly equal at 50 percent for TANF and 53 percent for food stamps.

Any Kind of Government Assistance. Receipt of publicly funded assistance broadly defined changed little between the first and second years following program entry, despite the steep reductions in TANF receipt. The rates of receipt of any kind of government assistance, as measured by the two follow-up surveys, were roughly stable in seven of the sites, increased by 9 percentage points in St. Lucie County, and fell by 3 to 13 percentage points in Baltimore County, Chicago, and Yakima (Exhibit V.4, top panel). The mean value of all assistance from government programs was also stable during this period, rising by less than 10 percent in two sites and remaining essentially unchanged in the others (Appendix Exhibit B.13).

The receipt of any kind of government assistance changed little primarily because of patterns in receipt of food stamps and SSI or SSDI. We have seen from the administrative data that, in most of the study sites, the receipt of food stamps by enrollees remained relatively stable while their receipt of TANF fell. In a few sites, the survey data show that increases in the receipt of SSI or SSDI partially offset reductions in the receipt of TANF.⁽⁵⁷⁾ This pattern is exemplified by Philadelphia, where the receipt of SSI or SSDI increased by 3 percentage points during the second year after program entry (Exhibit V.4, bottom panel). Ft. Worth, St. Lucie County, and West Virginia provide additional examples of this pattern.

3. Independence from TANF through Employment

PRWORA and the authorizing legislation for the WtW grants program both hold the objective of moving families off welfare and into employment. Exhibit V.5 presents evidence from the 24-month follow-up survey on the extent to which WtW enrollees achieved this objective. This exhibit shows the percentage distribution of enrollees in each study site across the four possible combinations of employment and receipt of TANF. The darkest section of the bar for each site shows the proportion of enrollees who were employed and off TANF. Not surprisingly, this proportion is largest, about two-thirds, for the two JHU sites, which targeted individuals who were employed when they entered WtW. In the other sites, about one-fourth to one-third of enrollees were employed and off TANF two years after program entry, except in West Virginia, where 44 percent had achieved this objective.

Despite the generally modest proportions of enrollees who achieved independence from TANF through employment, several sites saw some progress toward that objective during the second year. In four of the sites, the proportion of enrollees who were employed and off TANF was 4 to 6 percentage points higher at the end of the second year after program entry than at the end of the first year; however, the estimated increase is statistically significant only for the Chicago and Nashville sites (Exhibit V.6, top panel).

There was a pronounced increase during the second year following program entry in the proportion of WtW enrollees who were neither working nor receiving TANF. In seven of the study sites, this proportion increased by between 4 and 14 percentage points; and in five of these sites the estimated increase is statistically significant (Exhibit V.6, bottom panel). St. Lucie County experienced one of the largest increases, at 10 percentage points. This was also the site where enrollee receipt of SSI or SSDI increased most dramatically, by 7 percentage points, during the second year. The Ft. Worth and West Virginia sites also experienced substantial increases in the proportion of enrollees who were neither working nor receiving TANF, accompanied by increases in the receipt of SSI or SSDI. Thus, in three of the study sites, it appears that access to SSI or SSDI benefits facilitated an increase in the proportion of WtW enrollees who were neither employed nor on TANF.

B. HOUSEHOLD INCOME AND POVERTY

The household income of WtW enrollees was low or moderately low two years after program entry and was essentially unchanged from the previous year. Consequently, their poverty rates exceeded 50 percent in all of the study sites. Nevertheless, enrollees did not experience high levels of material distress during the second year following program entry — in several sites, an index of such distress actually fell during the second year.

1. Household Income

The mean monthly household income of WtW enrollees ranged from $1,100 to $1,800 at the end of the second year following program entry. In Baltimore County, Milwaukee, St. Lucie County, and Yakima, the mean household income was $1,500 per month or more (Exhibit V.7).⁽⁵⁸⁾ In these sites, the mean values of earnings by all household members, including the enrollee, were highest. In Baltimore County and St. Lucie County, the enrollees themselves generated most of the household earnings, as would be expected given the population targeted by the JHU programs in these sites. In sharp contrast, in the Milwaukee site, other persons contributed almost twice as much to household earnings as enrollees themselves. In Yakima, the enrollees and other members of their households contributed equal amounts of earnings.

WtW enrollees in Boston, Nashville, Philadelphia, and West Virginia had the lowest mean total household incomes — less than $1,200 per month. Earnings by the enrollees and others in their households were both low in these sites. These happen to be sites that followed a pre-employment model. Enrollees in sites that followed an employment model (Chicago, Ft. Worth, Phoenix, and Yakima) had higher mean total household incomes, but that was almost entirely due to higher earnings by other household members — a difference in outcomes that is almost certainly due to differences at baseline rather than to differences in the efficacy of the two program models.

Household income was remarkably stable between the end of the first year following program entry and the end of the second. There was no site for which this study found statistically significant evidence of a change in the mean household income (Appendix Exhibit B.20). The mean household income of enrollees in Milwaukee was about $300 per month higher at the end of the second year than at the end of the first, but given the site’s small sample size, this difference is not large enough to be statistically significant.⁽⁵⁹⁾

2. Poverty Rates

Most WtW enrollees were living in poverty two years after program entry. At least half of enrollees in every study site had household incomes that were below the federal poverty threshold (Exhibit V.8). More than two-thirds of enrollees in Boston, Chicago, Ft. Worth, Nashville, Philadelphia, Phoenix, West Virginia, and Yakima were in poverty at the two-year follow-up point. As a point of comparison, Loprest (2001) reports that in 1999, 52 percent of families nationwide that had left TANF in the previous two years were living in poverty.⁽⁶⁰⁾ In six of the WtW study sites, most enrollees were living in severe poverty, with household incomes below 50 percent of the federal poverty threshold. Rates of poverty and severe poverty, like mean household income, were essentially unchanged from the end of the first year after program entry to the end of the second (Appendix Exhibit B.20). These findings are based on a measure of income that does not include food stamps, the earned-income tax credit (EITC), income taxes, or payroll taxes.⁽⁶¹⁾

The incidence of poverty was much lower for enrollees who were employed at the end of the second year following program entry than those who were not.⁽⁶²⁾ Exhibit V.9 shows that the poverty rate was lower for employed enrollees in every study site by between 16 percentage points (West Virginia) and 43 percentage points (Baltimore County). However, even employed enrollees had high poverty rates; between one-half and three-fourths of the enrollees who were employed two years after entering WtW were living in poverty in all of the study sites except Baltimore County and Milwaukee, where about four in ten employed enrollees were in poverty. High poverty rates among employed enrollees reflect a combination of low wages and instability in employment during the month for which household income was measured.⁽⁶³⁾

3. Material Distress

A high incidence of poverty did not necessarily mean that WtW enrollees experienced especially high levels of material distress. This evaluation measured five types of material distress that enrollees and their families may have experienced during the second year after program entry: inability to fully pay the rent or mortgage, eviction, inability to fully pay a utility bill, termination of a utility, and disconnection of the telephone. Among these, inability to fully pay the rent or mortgage was most common in Milwaukee, Phoenix, and St. Lucie County, whereas inability to pay a utility bill in its entirety was most common in the other sites (Appendix Exhibit B.18). An index created for this evaluation summarizes the five types of material distress on a 0-to-1 scale, with higher values indicating greater distress.⁽⁶⁴⁾,⁽⁶⁵⁾ Across the study sites, the mean value of the index ranged from 0.20 in Baltimore County and West Virginia to 0.28 in Ft. Worth and Yakima, corresponding to the experience of between one and one-and-a-half types of distress (Exhibit V.10, upper panel).

Material distress among WtW enrollees fell in some sites between the first and second years following program entry. The mean value of the index of material distress was significantly lower in the second year in five study sites (Exhibit V.10, bottom panel). Thus, even though mean household incomes and poverty rates were unchanged in all of the study sites, material distress diminished modestly in some sites with the passage of time following program entry.

C. HARDSHIPS DURING THE SECOND YEAR

Although mean household incomes and poverty rates did not change between the first and second years following program entry, several specific hardships were less stable. This section presents evidence that homelessness, which was prevalent in some sites during both years, was somewhat less common during the second. On the other hand, enrollees’ lack of health insurance was rare during the first year but more common in most sites during the second.

1. Homelessness

WtW enrollees in three of the study sites — Boston, Milwaukee, and Phoenix — exhibited high rates of homelessness during the second year following program entry. Exhibit V.11 illustrates that 9 percent of enrollees in Boston lived in emergency or long-term shelters sometime during the year. A more extreme form of homelessness was common among enrollees in Milwaukee and Phoenix, where 12 percent and 10 percent, respectively, lived on the streets sometime during the year.⁽⁶⁶⁾ In sharp contrast, the rates of these two forms of homelessness did not exceed 2 percent in the predominantly rural sites of St. Lucie County and West Virginia. In the remaining study sites, rates were 7 percent or lower.

The incidence of homelessness fell during the second year after program entry in several study sites, particularly where rates of homelessness were high. The top panel of Exhibit V.13 shows that the percentage of enrollees who lived in a shelter fell significantly during the second year in Boston, Phoenix, and Yakima. The bottom panel shows that enrollees in Boston and Chicago were significantly less likely to have lived on the streets during the second year following program entry than during the first. Among the three sites with the highest rates of homelessness during the second year, rates had actually fallen in Boston and Phoenix but not in Milwaukee.

2. Lack of Health Insurance

WtW enrollees had more difficulty maintaining health insurance coverage for themselves than for their children. Exhibit V.12 shows near-universal coverage (93 percent or more) by public or private health insurance for the children of WtW enrollees two years after program entry in all but the Ft. Worth (84 percent) and St. Lucie County (89 percent) sites. The coverage rates for children were essentially unchanged from the previous year, as illustrated in the top panel of Exhibit V.14. For the enrollees themselves, the median coverage rate across the study sites was just 75 percent. The enrollee coverage rate was lowest in Milwaukee, where very low rates of TANF receipt (4 percent, as shown in Exhibit V.2) by the noncustodial parent enrollees reduced the likelihood they would be covered by Medicaid (Exhibit V.12). In six sites, the percentage of enrollees covered by health insurance fell during the second year following program entry by between 4 points (Philadelphia) and 12 points (West Virginia), as shown in the bottom panel of Exhibit V.14.

Loss of Medicaid coverage following exit from TANF appears to have been a factor behind the reductions in health insurance coverage for WtW enrollees. In five of the six sites where enrollee health insurance coverage rates fell the second year after program entry, rates of receipt of TANF also fell. A comparison of the bottom panel of Exhibit V.14 with Appendix Exhibit B.12 reveals that rates of health insurance coverage and TANF receipt both fell significantly in Boston, Chicago, Ft. Worth, Philadelphia, and West Virginia. St. Lucie County was the only site where the health insurance coverage rate for WtW enrollees fell but there was no significant reduction in the rate of TANF receipt. Thus, it appears that exit from TANF exposed enrollees in about half of the study sites to a greater risk that they, but not their children, would be without health insurance coverage.⁽⁶⁷⁾

[ Go to Contents ]

Endnotes

(50) In Milwaukee and Philadelphia, where rates of receipt of assistance from community organizations were especially low during the first year after program entry, WtW enrollees experienced little or no erosion in such support during the second year.

(51) The government assistance programs considered in this study are: TANF, food stamps, SSI or SSDI, Social Security retirement or survivors’ benefits, unemployment insurance benefits, general relief or general assistance, foster care or adoption assistance, and other assistance (including workers’ compensation and veterans’ benefits but not housing or medical assistance). Appendix Exhibit B.11 presents findings on the receipt and amounts of these types of assistance.

(52) The discussion of TANF and food stamp receipt is based on data for early WtW enrollees rather than all enrollees. This is because in 10 of the study sites, a full 8 quarters of post-entry data are available only for early enrollees, whereas only six or fewer quarters of post-entry data are available for all enrollees. Early enrollees are defined as those who entered WtW prior to July 1, 2000. Chicago is the only study site in which there were no early enrollees; consequently, it is the one site for which less than 8 quarters of post-entry administrative data are available. The 4 quarters of post-entry results shown in Exhibit V.3 for Chicago are based on all WtW enrollees. Among all of the study sites, in those quarters for which administrative data are available for all enrollees as well as early enrollees, quarter-specific TANF participation rates are generally quite similar between the two groups, as shown in Appendix Exhibits D.2.a and D.2.b. This is especially true for the latest of those quarters, typically Quarters 4 through 6. Thus, we believe that our decision to focus on results for early enrollees does not distort the conclusions that we draw from these data.

(53) The relative change in the rate of TANF receipt over the two years following WtW program entry is defined as the change in the rate of receipt between Quarter 0 and Quarter 8, divided by the rate of receipt in Quarter 0 and expressed as a percentage. All percentage changes reported in this section have been calculated in this manner.

(54) The results for Chicago are based on the four quarters of post-entry administrative data that were available to this study.

(55) State administrative data for all families receiving TANF in the study sites were not available for Boston and Milwaukee.

(56) From fiscal year 2000 to June 2003, the average monthly number of individuals receiving TANF nationwide fell by 17 percent (Committee on Ways and Means of the U.S. House of Representatives 2004; DHHS 2004) while the number receiving food stamps rose by 27 percent (USDA 2004). We would not expect patterns of receipt by cohorts of individuals or families, such as those who participated in the WtW evaluation, to mirror these caseload trends. However, the caseload trends do provide supportive context for our finding that the receipt of TANF by WtW enrollees fell relative to the receipt of food stamps.

(57) This is consistent with the evaluation team’s observation site visits to local TANF offices under this and other studies that many TANF agencies were making efforts to identify individuals on their caseloads who were close to exhausting their TANF five-year limit and might qualify for SSI or SSDI and to help them with their applications.

(58) The measure of income reported here is for the month prior to the month in which the 24-month survey interview was completed. It includes cash income of all types (the enrollee’s own earnings, earnings of other household members, cash benefits from government programs, and any other cash income), plus food stamps, received by all members of the enrollee’s household. See Appendix Exhibit B.16 for additional details on the calculation of total household income.

(59) The study sample size in Milwaukee was 276, of whom 195 responded to the 12-month follow-up survey and 190 responded to the 24-month follow-up survey. See Appendix F of this report and Appendix C of Fraker et al. 2004 for additional details on this study’s survey data collection.

(60) Loprest (2001) based her analysis on the 1999 wave of the National Survey of America’s Families. To be included in that analysis, families had to have left TANF sometime during the two years prior to the survey interview and to have not been receiving TANF at the time of the interview. Seventy-nine percent of these families included at least one parent who was employed at the time of the interview.

(61) To be consistent with the Census Bureau’s methodology for determining poverty status, we excluded food stamps from the previously-described measure of income (see footnote 58) for the purpose of the poverty analysis. For the same reason, we did not factor payroll taxes, income taxes, or the EITC into the poverty analysis.

(62) Elsewhere in this report, employment is defined as working on a job for pay at the time of the survey interview. However, to obtain a consistent reference period for earnings and employment in the analysis of poverty by employment status, “employment” here is defined as positive enrollee earnings in the month prior to the survey interview (the month for which income data were reported). This definition ensures that there are no “employed” enrollees without earnings in the analysis of poverty by employment status. If such individuals had been included in the analysis, they would have tended to exaggerate the poverty rate among employed enrollees.

(63) We simulated household poverty rates for employed enrollees under the assumptions of: (1) no income from government programs (including the EITC), (2) the enrollee’s actual earnings in the month prior to the month of the survey interview, and (3) whatever other income their households actually received in that month. The simulated poverty rates are as follows (in percentages): Baltimore County 51, Boston 60, Chicago 74, Ft. Worth 62, Milwaukee 41, Nashville 74, Philadelphia 81, Phoenix 63, St. Lucie County 62, West Virginia 81, and Yakima 60. We then repeated the simulation, but replaced the enrollee’s actual earnings with earnings calculated on the basis of the usual weekly hours of work on the principal job held at the time of the survey interview, the hourly wage on that job, and 4.3 weeks of work in the month. Poverty rates under the second simulation are less than or equal to those under the first simulation; typically, they are about 10 percentage points lower: Baltimore County 34, Boston 41, Chicago 66, Ft. Worth 56, Milwaukee 41, Nashville 62, Philadelphia 65, Phoenix 48, St. Lucie County 52, West Virginia 75, and Yakima 51. An important factor contributing to the higher simulated poverty rates based on actual earnings is lack of consistent work by employed WtW enrollees over the month preceding the survey interview. We conducted these simulations using data for enrollees who (1) had positive earnings in the month prior to the survey interview and (2) had positive hours of work and a positive wage rate on the principal job held at the time of the interview.

(64) The value of the index of material distress was computed by adding up the number of an enrollee’s affirmative responses to questions regarding the presence of the five types of material distress and dividing by the number of valid responses. If all five types of distress were experienced, the index took on its maximum value of 1; if only one type was experienced, it took on a value of 0.2 (assuming valid responses to all five questions); and if no type of distress was experienced, the index took on its minimum value of 0.

(65) The design of the index of material distress closely resembles that of the “index of material hardship” used in several random assignment evaluations of state welfare reform initiatives in the 1990s (Bloom et al. 2002, Fraker et al. 2002, and Miller et al. 2000). For this study, we omitted two of seven specific types of distress, both reflecting failure to see a health care professional when needed. We included these instead in this study’s index of health-related distress (Appendix Exhibit B.22).

(66) These three study sites were the only ones in which the incidence of either form of homelessness among WtW enrollees exceeded 10 percent during the second year following program entry (Appendix Exhibit B.18).

(67) As noted in Section C.3 of Chapter IV, coverage by employer-provided health insurance increased among employed enrollees between the end of the first year following program entry and the end of the second year in Chicago, Philadelphia, and Nashville. Rates of TANF receipt, and presumably rates of own Medicaid coverage, fell among all enrollees in these same sites over that period. These two trends had opposing implications for enrollee health insurance coverage. The net effect was a reduction in health insurance coverage among enrollees in Chicago and Philadelphia and essentially no change in coverage among enrollees in Nashville.

Where to?

Top of Page | Contents

Main Page of Report | Contents of Report

Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation ASPE)
U.S. Department of Health and Human Services (HHS)