Table of Contents | Previous | Next |
Summary Report
The Family Transition Program (FTP) was a welfare reform pilot project that operated from 1994 to 1999 in Escambia County, Florida — a mid-sized county that includes the City of Pensacola. FTP was one of the first welfare reform initiatives to impose a time limit on the receipt of cash assistance — 24 months in any 60-month period for most recipients and 36 months in any 72-month period for the least job-ready — and was the first program in the nation in which families reached a time limit and had their welfare benefits canceled. In addition to its time limit, FTP included an unusually rich array of services, mandates, and financial work incentives designed to help welfare recipients prepare for, find, and hold jobs.
FTP was implemented more than two years before the passage of the 1996 federal Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA), and it anticipated key elements of the federal law. FTP also served as a pilot for Florida’s statewide welfare reform program, implemented in late 1996. Thus, FTP provides important lessons on the implementation and potential effects of more recent welfare reform initiatives in Florida and elsewhere — although this evaluation does not measure the effectiveness of Florida’s current statewide welfare program.
In 1994, the Florida Department of Children and Families contracted with the Manpower Demonstration Research Corporation (MDRC) to conduct a multifaceted evaluation of FTP’s effectiveness. The study was also funded by the U.S. Department of Health and Human Services and the Ford Foundation, and its analysis of FTP’s effects on children was funded by the agencies and foundations listed at the front of this report. MDRC is a nonprofit, nonpartisan organization with 25 years’ experience designing and evaluating social policy initiatives, including many state and federal welfare reforms.
To assess what difference FTP made, the study compared the experiences of two groups of people: the FTP group, which was subject to the program, and the Aid to Families with Dependent Children (AFDC) group, which was subject to the prior welfare rules (including, for many recipients, a requirement to participate in employment-related activities through Project Independence, Florida’s pre-existing welfare-to-work program). To ensure that the groups would be comparable, welfare applicants and recipients were assigned at random to one or the other group. Because the two groups had similar kinds of people, any differences that emerged between the groups during the study’s follow-up period can reliably be attributed to FTP rather than to differences in personal characteristics or changes in the external environment.
This is the fifth and final report in the FTP evaluation. It summarizes the earlier findings and provides new information in several areas. It follows eligible families for at least four years after they entered the study — well beyond the point when recipients began reaching the time limit — and uses data from a large-scale survey to assess, for the first time, FTP’s effects on key outcomes such as food security and child well-being. In addition, the report provides new information from in-depth, post-welfare interviews with FTP participants whose benefits were canceled at the time limit. Finally, the report describes the results of a benefit-cost analysis, which compares FTP’s financial benefits and costs for participants and government budgets.
I. Background: FTP and the Evaluation
A. The Family Transition Program
The Family Transition Program was created by the Florida legislature in April 1993 and began operating in February 1994 under waivers of federal welfare rules.1 (These waivers were no longer needed after 1996 because FTP’s provisions are permitted under the federal welfare law.)
FTP tested a model that combined a time limit on cash assistance receipt with an array of services and supports designed to help participants prepare for, find, and hold jobs. Its main goals were to increase self-sufficiency and reduce long-term welfare dependency. The key features, summarized in Table 1, included:
-
A time limit. Most FTP participants were limited to 24 months of cash assistance receipt in any 60-month period.2 Certain groups were exempt from the time limit, and, in addition, the program policies included several safeguards that could, in theory, lead to temporary benefit extensions for families reaching the time limit, partial (rather than full) benefit termination, or post-time-limit subsidized jobs. The AFDC group was not subject to a time limit.
-
Financial work incentives. Under FTP, the first $200 plus one-half of any remaining earned income was disregarded (that is, not counted) in calculating a family’s monthly grant. Known as an enhanced earned income disregard, this policy allowed a greater proportion of working families to retain at least a partial welfare grant. Although FTP’s disregard was generous, its effect on recipients’ income was limited by Florida’s relatively low welfare grant levels (a maximum of $303 for a family of three): A mother with two children working half-time at the minimum wage had about $100 more income per month under FTP than under AFDC. In addition to the enhanced disregard, FTP allowed families to accumulate more assets and to own more valuable cars (relative to AFDC rules) without losing eligibility for welfare. Finally, FTP offered subsidized transitional child care for two years after participants left welfare for work, as opposed to the one year provided under prior rules.
-
Enhanced services and requirements. FTP aimed to provide a rich array of services and supports. Most notably, participants received intensive case management provided by workers with very small caseloads. FTP participants were also more likely than AFDC group members to be required to participate in employment-related activities, and the program developed some enhanced education, training, and job placement services. Finally, FTP sought to increase participants’ access to a range of other benefits, including social and health services, child care, transportation, and other support services by increasing funding for such services and bringing many of them under one roof in the program offices.
-
Parental responsibility mandates. Under FTP, parents with school-age children were required to ensure that their children were attending school regularly and to speak with their children’s teachers at least once each grading period. New applicants for welfare who had preschool children were required to provide proof that their children had begun to receive the standard series of immunizations. None of these mandates existed for the AFDC group. Parents who failed to meet these requirements — as well as those who did not comply with the employment and training participation mandates described above — faced sanctions (that is, their grants could be canceled or reduced).
| Characteristic | FTP Policy | AFDC Policy |
|---|---|---|
| Time limit on cash assistance receipt | 24 months in any 60-month period for most recipients; 36 months in any 72-month period for the least job-ready. Exceptions under certain circumstances. | None |
| Amount of earned income disregarded in calculating cash assistance grants | The first $200 plus 50% of any remaining earnings. | First 4 months of work: $120 plus 33% of earnings; |
| Months 5-12: $120 disregarded; | ||
| After month 12: $90 disregarded. | ||
| Asset limit for cash assistance eligibility | $5,000 | $1,000 |
| Value of vehicle excluded in counting assets for cash assistance eligibility | $8,150 | $1,500 |
| Child care assistance for families leaving welfare for work | Two years of transitional child care assistance; eligibility beyond that point depended on eligibility for other programs. | One year of transitional child care assistance; eligibility beyond that point depended on eligibility for other programs. |
| Exemptions from employment-related mandates for recipients with young children | Parent exempt if caring for a child under 6 months old. | Parent exempt if caring for a child under 3 years old. |
| Parental responsibility mandates | Parents had to ensure that children attended school regularly, and had to speak with teachers at least once each grading period. Applicants with preschool children had to prove that children had begun immunizations. | None |
| Employment-related, social, and health services | Participants received intensive case management and a range of social and health services; enhanced employment related services. | Participants were served by the pre-existing Project Independence welfare-to-work program. |
B. FTP’s Policy Significance
Although the 1996 federal welfare law fundamentally changed the structure and funding of cash assistance for needy families, many of the specific policies that the law encourages states to adopt were already being implemented under waivers of federal AFDC rules that were granted to 43 states prior to the bill’s passage. For example, more than 30 states received waivers to implement some form of time limit on welfare receipt in at least part of the state. The federal law replaced AFDC with the Temporary Assistance for Needy Familes (TANF) block grant, and it restricted states from using federal TANF funds to provide assistance to most families for more than 60 months. Although states may exempt up to 20 percent of the caseload from this provision, they also may set time limits of fewer than 60 months.
FTP was one of the most important initiatives implemented under waivers because it was one of the first to include a time limit. Time limits have been among the most controversial features of state and federal welfare reform efforts in the 1990s. Proponents argue that time limits are necessary to send a firm message to recipients (and the system) that welfare should be temporary; they maintain that the limits will motivate recipients to find jobs or other means of support for their families. Critics contend that many recipients face serious personal problems or skills deficits that make it difficult for them to support their families for long periods without assistance; thus, they argue, time limits will cause harm to many vulnerable families.
Although time limits have been in place in a few areas for as much as six years, there are still relatively few data available to inform this debate. Overall, 25 states (including the District of Columbia) have imposed a 60-month time limit, and no families have reached those limits yet. Another eight states have not imposed time limits that result in cancellation of families’ entire welfare grants.3 Together, these two groups of states account for about three-fourths of the national welfare caseload.
On the other side of the spectrum, 17 states — accounting for about one-fourth of the national caseload — have imposed time limits that could result in cancellation of a family’s entire grant after fewer than 60 months of welfare receipt. Even among these states, however, the specific rules and their implementation vary widely. In some states, a large proportion of the welfare caseload is exempt from the time limit. Other states have granted extensions to many of the families who have reached the time limits. As a result, there are only a handful of states in which a substantial number of families have had their benefits canceled at a time limit. A few of these states are tracking the families whose cases were closed, and an even smaller number are sponsoring random assignment evaluations that will provide reliable information on program effects.
In short, while the FTP evaluation is not designed to isolate the impact of the time limit per se — the program was an integrated package of services, incentives, and time limits — the study is one of only a few sources of reliable evidence on the implementation and effects of one the most important recent changes in welfare policy.
In Florida, FTP was the precursor to WAGES (Work and Gain Economic Self-Sufficiency), a statewide welfare reform that operated from 1996 to 2000. FTP and WAGES shared many features, including the time limit, enhanced earned income disregard, and extended transitional child care.4 At the same time, while the implementation of WAGES varied across the state, it generally did not include FTP’s emphasis on very intensive services and case management. In 2000, WAGES was merged with the state’s workforce program, but many of the key policies (including the time limit) remain in place.
C. The FTP Evaluation
The FTP evaluation, which began in early 1994, was initially required as a condition of the federal waivers that allowed Florida to implement the program. The state elected to continue the evaluation even though it was not required to do so under the 1996 federal welfare law.
The evaluation includes three major components:
-
Implementation analysis. This part of the study examines how FTP operated. Data on a program’s implementation can be critical to interpreting its impacts and to identifying practices that are associated with success.
-
Impact analysis. This part of the study assesses whether FTP generated changes in participants’ employment, earnings, welfare receipt, family income, and other outcomes, relative to the AFDC system it replaced. The impact analysis is also examining FTP’s effects on families and children.
-
Benefit-cost analysis. This analysis uses data from the impact analysis and from agency fiscal records to compare the financial benefits and costs of FTP for both the government budget and families subject to the program.
As noted earlier, the impact analysis was based on a random assignment research design. Although this design has some limitations — for example, the study cannot assess whether FTP affected the number of people who initially applied for welfare — random assignment is generally considered to be the most reliable way to determine what difference, if any, a program makes.
People were assigned to the FTP and AFDC groups when they applied for welfare or, if they were already receiving benefits, when they came to the welfare office for a recertification interview. Three key aspects of this process are worth noting:
-
Certain groups of recipients — including those who asserted that they were incapacitated and unable to work — were screened out prior to random assignment and did not enter the study.5 Thus, the study does not provide information on the impact of FTP for the full welfare caseload — including, potentially, a small but very hard-to-employ segment of the population. (As discussed below, some other participants were exempted from FTP after they were randomly assigned; they remained in the study.)
-
Welfare applicants were randomly assigned before staff knew whether their application would be approved. Thus, around 8 percent of the FTP group never received cash assistance during the follow-up period, either because they did not follow through with their application or because they were found to be ineligible for benefits. These individuals had little or no contact with the program.
-
Unlike many earlier studies, this one did not compare FTP with a control group that was not required to engage in any employment-related activities. In accordance with prior rules, many members of the AFDC group were required to participate in Project Independence (PI). As a result, the impact analysis assessed what difference FTP made above and beyond the impact produced by AFDC/PI.
The evaluation focused on the approximately 2,800 single parents (1,400 in each group) who were randomly assigned to the FTP and AFDC groups from May 1994 (when FTP began full-scale operations) to February 1995; these individuals are known as the report sample. Thus, the evaluation included mostly people who entered FTP during its start-up period.
Almost all of the report sample members are women, and their average age was about 29 years old when they entered the study. Although most had small families, about two-thirds had at least one preschool child, and more than 40 percent had a child under 2 years old. Roughly equal proportions of the sample are black and white; there are few Hispanics. The vast majority of sample members had at least some work experience, but most had little recent work experience, and 40 percent had never worked full time for six months or more for one employer. Nearly 40 percent did not have a high school diploma or equivalent. About half were applying for welfare when they were randomly assigned, but only 12 percent were first-time applicants; more than half had received welfare for a total of two years or more prior to random assignment.
The study used a variety of data sources to assess FTP’s implementation and impacts. Key among these were administrative records of sample members’ monthly cash assistance and Food Stamp benefits in Florida, quarterly earnings in jobs covered by Florida’s Unemployment Insurance (UI) system, child care subsidy payments, and Medicaid-covered health expenditures.
In addition, the study drew on two relatively large-scale surveys of FTP and AFDC group members. The first, administered about two years after people were randomly assigned, included about 600 respondents (300 in each group) and was mainly used to assess FTP’s implementation and its program message. The second survey was administered to more than 1,700 people (a little more than 850 in each group) roughly four years after random assignment.6 More than 1,100 of those who responded to the four-year survey — those with at least one child between 5 and 12 years old when interviewed — answered a special 90-minute segment of questions about child care, parenting, and child well-being. Both surveys achieved high response rates: 80 percent of targeted clients were located and interviewed.
Finally, MDRC examined the implementation of both FTP and AFDC/PI by interviewing staff, observing program activities, reviewing client case files, administering a staff survey, and holding focus groups with participants. The cost analysis drew on a variety of fiscal reports and other program records.
D. The Context
In considering the broader applicability of the FTP experience, it is critical to understand the unusual context in which the program operated. Three factors are particularly important:
-
Socioeconomic conditions. Escambia is a mid-sized county with no large cities; the local unemployment rate was at or below the already-low state and national rates throughout the study period.
-
Welfare reform environment. FTP was implemented during a period of extraordinary change in state and federal welfare policy. The federal welfare law and Florida’s statewide welfare reform were both enacted about two years after FTP began operating. In addition, Florida’s welfare caseload declined at an unprecedented rate during the period. After more than doubling from 1989 to late 1993, the caseload plunged by 71 percent from January 1994 to June 1999. There is no doubt that the AFDC group was affected to some extent by the broad public discourse about welfare reform.
-
Timing. FTP was implemented when time limits were still a new and unfamiliar concept. Many participants (and some staff) initially expressed uncertainty or skepticism about whether families’ benefits would actually be terminated at the time limit.
Together, these factors suggest that the evaluation represents a conservative test of FTP’s impacts — that the measured impacts might have been larger if the AFDC group had been completely unaffected by welfare reform and if FTP had not been the first program of its type.
Nevertheless, the weight of the evidence suggests that FTP received a fair test. The data presented below show that the FTP and AFDC groups had dramatically different experiences while on welfare. FTP sent a sharply different message and provided different services than AFDC/PI, and its time limit was real. If these key program components truly affected participants’ outcomes, this would be reflected in program impacts.
II. Evaluation Results
A. FTP’s Implementation
Ultimately, FTP provided an impressive array of services and supports for participants. Each participant was assigned to a case manager and an employment and training worker; the two types of workers were stationed in the same office and had overlapping caseloads to facilitate communication. In addition, the FTP offices housed computerized learning labs and a variety of outstationed staff from other agencies (for example, a child care counselor, a mental health worker, and a nurse). The program was hindered at various points by staff turnover, difficulties with interagency linkages, and other issues, but it still looked dramatically different from AFDC.
It is important to note, however, that FTP began operating just three months after Escambia was selected as a pilot county; thus, local planners had little time to assemble the enhanced model. As a result, some pieces of the service package were not in place when participants began to enroll, and some early enrollees did not receive a fully implemented version of FTP. This further supports the conclusion that the study results are a conservative estimate of FTP’s potential.
Nevertheless, data from surveys and interviews with staff and clients indicate that, even within the report sample, the FTP group had quite different experiences than the AFDC group. For example:
-
As shown in Figure 1, the FTP group was substantially more likely to participate in employment-related activities. This occurred in part because AFDC group members were not required to participate if they had a child under 3 years old (FTP exempted only those with a child under 6 months old). In addition, while both groups received the same general types of employment services, FTP developed enhanced services in several areas (for example, special compressed vocational training programs). FTP was not a strict “work first” program in which job search and quick employment are strongly emphasized; it increased participation in both job search activities and education and training. The program also increased the number of people who obtained a trade license (not shown in the figure).
-
FTP case managers had very small caseloads (typically around 35 active cases per worker), allowing them to deliver more personalized services than their counterparts who worked with the AFDC group. In addition, FTP staff transmitted a message focusing more heavily on self-sufficiency. Figure 2, drawn from the two-year client survey, shows that FTP group members were more likely to report that staff knew about them and their situations and that they heard a different message while on welfare. Finally, FTP participants were much more likely to be sanctioned for failing to follow program rules, at least in the early part of the follow-up period (not shown in the figure).
-
Figure 2 also shows that FTP staff did a good job of informing participants about the time limit. However, the program’s message, at least in the early operational period, focused more on skill-building to prepare for “good” jobs and less on leaving welfare quickly to “bank” available months. The figure also shows that some members of the AFDC group believed, erroneously, that they were subject to a time limit.
[D] |
Despite all of FTP’s expanded services and supports, Figure 2 shows that, on the two-year client survey, FTP participants were only slightly more likely than AFDC group members to agree with the statement “I received help that improved my long-term chances of getting or keeping a job.”7
B. The Time Limit
Escambia County was the first place in the United States where families reached a welfare time limit and had their benefits canceled; the first families reached the limit in 1996. Key findings related to the time limit include:
- More than three-fourths of the FTP group received benefits for less than the 24 or 36 months allowed under their time limit.
About 55 percent of the FTP group was subject to a 24-month time limit. Of this group, only 16 percent accumulated 24 or more months of benefit receipt with four years after entering the study. Among the least job-ready participants — those subject to a 36-month time limit — 27 percent received at least 36 months of benefits within four years. Thus, overall, about 21 percent of the FTP group received at least as many months of benefits as their time limit allowed; the others left welfare before reaching that point (some cycled off and back onto welfare, but still did not accumulate 24 or 36 months of benefits by the end of the study period).
[D] |
- About two-thirds of those who received 24 or 36 months of benefits — one-sixth of all FTP participants — had their welfare grants canceled owing to the time limit.
FTP’s rules included several safeguards related to the time limit. First, participants could be exempted if a physician found them to be incapacitated; their time-limit clock was suspended while the exemption applied (as noted earlier, people who were known to be incapacitated at the outset did not enter the program or the study).8 Second, participants who reached the time limit could receive up to two four-month benefit extensions if they had “substantially complied with their FTP plan” but encountered “extraordinary difficulties” in finding a job or completing their assigned activities. Third, if full benefit termination was deemed “likely to result in a child’s being placed into emergency shelter or foster care,” the children’s portion of the benefit was to be continued and diverted to a third party to administer on their behalf.
Finally, under terms of the federal waiver, Florida was required to provide a public or private transitional work opportunity to “each FTP participant who has diligently completed her self-sufficiency plan but has been unable to find employment at the end of the . . . time limit.” The waiver required the state to provide a public job if a private job could not be found.9
FTP developed a complex, multistep process to review cases approaching the time limit, in order to determine when the various safeguards should be applied. The process included an unusual entity known as a Review Panel, which was composed of volunteers from the community. Despite the many safeguards and layers of review, however, only the first of the policies (exemptions) was used in a significant number of cases.
As shown in Figure 3, by June 1999 (shortly before FTP ended), a total of 340 members of the report sample had accumulated at least as many months of benefit receipt as their time limit allowed (that is, 24 months of receipt if they were subject to a 24-month limit, and 36 months if they were subject to a 36-month limit).10 Of this group, 103 never reached the time limit, however, because some of their months of benefit receipt were not counted — usually because they received a medical exemption that stopped their time-limit clock (a few moved to other Florida counties, which initially did not have time limits). Thus, a total of 237 people — 17 percent of the report sample — actually reached the time limit.
The bottom section of the figure shows that, of the 237 sample members who reached the time limit, 227 (96 percent) had their welfare grant fully canceled (a handful received a brief extension before their grant was canceled). In the other cases, the children’s portion of the grant was retained. No one was given a post-time-limit transitional job.
[D] |
Two factors explain the small number of extensions and the absence of transitional jobs. First, according to program records, nearly 40 percent of those who reached the time limit were already employed and earning at least as much as a standard welfare grant plus $90 (the program referred to this as “grant plus $90”). These participants were considered self-sufficient and not in need of an extension or a transitional job.11 (In fact, many of these participants would have become ineligible for welfare before reaching the time limit had it not been for FTP’s enhanced earned income disregard.)
Second, the vast majority of the people who reached the time limit without a job paying at least grant plus $90 were deemed to have been noncompliant with FTP, a designation that made them ineligible for a transitional job and very unlikely to receive an extension. “Noncompliance” was never precisely defined, and interviews with staff suggested that the distinction between failure to follow program rules and failure to make progress toward self-sufficiency became blurred in practice.
- The FTP participants who reached the time limit were a diverse group and were not necessarily the most disadvantaged participants.
In comparison with other FTP group members, those who reached the time limit were more likely to have received large amounts of welfare before entering FTP, to have very young children, and to be African-American. Nevertheless, even among these groups, most did not reach the time limit. For example, among those who had received welfare for five years or more prior to enrollment, only 22 percent reached the time limit. It appears that some of the participants facing the most serious barriers to employment (for example, health or emotional problems) were granted exemptions and thus did not reach the time limit.
In addition, the group reaching the time limit was far from homogeneous. For example, while half had a child under 2 years old at enrollment, one-fourth had no preschool children. In addition, they had different experiences while in FTP. More than three-quarters worked in the year prior to reaching the time limit (mixing work and welfare), and more than one-fourth worked throughout that year. As noted earlier, many of these participants presumably would have left welfare earlier had it not been for FTP’s enhanced earned income disregard. In-depth interviews suggest that some of those who did not work in the pre-time-limit period faced serious barriers to employment; others were being supported by their parents or partners and may have felt little urgency about finding a job; and still others were attending post-secondary education or training programs while in FTP (with or without the program’s consent).
C. FTP’s Impacts on Employment, Public Assistance Receipt, and Other Economic Outcomes
The main impact analysis followed about 1,400 people in each research group for four and a half years after each person’s random assignment date (for simplicity, most measures include only the first four years of follow-up). Administrative records of cash assistance receipt (referred to as AFDC/TANF), Food Stamp receipt, and quarterly earnings in UI-covered jobs were available for all sample members. Outcomes such as job characteristics, material hardship, and health coverage were examined using survey data, which were available for just over 850 people in each group who responded to the four-year client survey. Key findings on economic outcomes include:
- On average, over the four-year follow-up period, FTP increased employment and earnings, reduced welfare receipt, and modestly raised participants’ income.
Table 2 summarizes FTP’s impacts on employment and public assistance outcomes over the entire four-year follow-up period. These data are drawn from administrative records.
As is clear from the table, the AFDC group left welfare very quickly. Only 17 percent accumulated more than 36 months of cash assistance (AFDC/TANF) during the four-year period. Although not shown in the table, about 96 percent of the AFDC group left welfare, at least temporarily. This reflects the rapid overall decline in Florida’s welfare caseload during this time.
Nevertheless, FTP still reduced cash assistance receipt: Only 6 percent of the FTP group received benefits for more than 36 months. Over the entire period, the FTP group received an average of $3,987 in cash assistance, roughly $700 (15 percent) less than the AFDC group average. As discussed below, these impacts appear to have been due largely to the time limit. FTP also reduced Food Stamp payments by about $500 per person (8 percent), although it did not affect the rate of Food Stamp receipt. The asterisks in Table 2 indicate that these differences are statistically significant, meaning that they are unlikely to be due to chance.
The AFDC group was also quite likely to work. Table 2 shows that 82 percent worked in a UI-covered job at some point. FTP did not increase the number of people who ever worked, but it did increase the amount that people worked. As the table shows, the average quarterly employment rate was about 48 percent for the FTP group and 44 percent for the AFDC group. As a result, average earnings over the full period were about $2,400 (17 percent) higher for the FTP group.
In dollar terms, the FTP group gained about twice as much in earnings as they lost in public assistance. Thus, Table 2 shows that members of the FTP group had nearly $1,200 more in combined income from these sources over the entire follow-up period, and they also derived a greater share of income from earnings and a smaller share from public assistance. The magnitude of the income gain was modest, however — the FTP group had about $300 more income per year, on average. It is important to note that this is not a complete measure of household income, because it does not include sample members’ income from other sources (for example, child support and the federal Earned Income Credit)12 or the income of other household members.
- The pattern of FTP’s impacts on employment, welfare receipt, and income shifted significantly over the four-year follow-up period.
| Outcome | FTP Group |
AFDC Group |
Difference (Impact) |
Percentage Change |
|---|---|---|---|---|
| Employment | ||||
| Ever employed (%) | 84.1 | 82.4 | 1.8 | 2.1 |
| Average quarterly employment rate (%) | 48.3 | 43.8 | 4.5*** | 10.3 |
| Public assistance receipt | ||||
| Average months receiving AFDC/TANF | 15.4 | 17.1 | -1.7*** | -9.9 |
| Received more than 36 months of AFDC/TANF (%) | 6.1 | 16.5 | -10.4*** | -62.8 |
| Average months receiving Food Stamps | 24.6 | 24.8 | -0.2 | -0.9 |
| Income from earnings and public assistance | ||||
| Average total earnings ($) | 16,666 | 14,288 | 2,378*** | 16.6 |
| Average total AFDC/TANF benefits ($) | 3,987 | 4,698 | -711*** | -15.1 |
| Average total Food Stamp benefits ($) | 6,121 | 6,621 | -499*** | -7.5 |
| Combined
income from earnings, AFDC/TANF, and Food Stamps ($)a |
26,774 | 25,606 | 1,167* | 4.6 |
| At least 50 percent of income from earnings (%) | 50.1 | 44.7 | 5.4*** | 12.1 |
| Sample size | 1,405 | 1,410 | ||
| SOURCES: MDRC calculations from Florida Unemployment
Insurance (UI) earnings records, AFDC/TANF records, and Food Stamp
records. NOTES: A two-tailed t-test was
applied to differences between the FTP and AFDC groups. Statistical
significance levels are indicated as *** = 1 percent; ** = 5 percent;
* = 10 percent. Rounding may cause slight discrepancies in the calculation of sums and differences. |
The top two panels of Figure 4 illustrate the pattern of FTP’s impacts on earnings and AFDC/TANF payments over the entire follow-up period. The top panel shows that FTP’s impact on earnings emerged early in the follow-up period, peaked in years 2 and 3, and then disappeared by the end of year 4. At the end of the period, the employment rates for the two groups (not shown in the figure) were nearly identical. Much of the decay in FTP’s impact on employment and earnings occurred because the AFDC group “caught up” to the FTP group in year 4. For example, among those not employed at the end of year 3, AFDC group members were more likely than their FTP group counterparts to work during year 4 (not shown). It is possible that the statewide implementation of WAGES — and the accompanying heavy publicity — affected the behavior of some AFDC group members, even though those who remained in Escambia County were not actually subject to WAGES until after the study ended.
The middle panel of Figure 4 shows that the impacts on cash assistance payments exhibited a somewhat different pattern. FTP did not reduce the rate of cash assistance receipt in the first two years of follow-up, before anyone reached the time limit (although, as shown in the figure, FTP did begin to reduce welfare payment amounts during year 2). Both groups left welfare rapidly, and the program’s main impact during this period was to increase significantly the number of people combining work and welfare. One would normally expect an enhanced earnings disregard such as FTP’s to increase the number of people on welfare. The fact that FTP did not increase cash assistance receipt implies that the program may have generated offsetting effects — some elements of the program (for example, strong participation mandates and the impending time limit) may have induced participants to leave welfare more quickly in the pre-time-limit period, while other elements (for example, the enhanced disregard) induced people to stay on welfare longer. These effects could have worked in opposite directions, resulting in no impact overall.13
The pattern of impacts on welfare receipt changed abruptly when FTP participants began reaching the time limit: The program reduced the number of people receiving cash assistance throughout years 3 and 4 and, as shown in Figure 4, the impact on cash assistance payments grew larger.
The pattern of income impacts follows from the earnings and welfare results discussed above. The bottom panel of Figure 4 shows that income gains were concentrated in year 2 and year 3, when the FTP group’s earnings gains were more than large enough to offset their lower public assistance amounts. By the end of the follow-up period, however, the earnings gains had diminished and were about equal in dollar terms to the losses in public assistance. As a result, the positive impact on total income disappeared. The decline in income impacts does not erase the income gains that occurred earlier in the follow-up period, but it strongly suggests that the FTP group will not accumulate additional income gains relative to the AFDC group over time.
Figure 5 illustrates the impact trends in a different way, showing the average amount of earnings, AFDC/TANF, and Food Stamps for each research group in each year of the follow-up period — and, at the top of each bar, the sum of the three income sources. Figure 5 clearly shows that both research groups relied increasingly on earnings and less on public assistance over time. Nevertheless, particularly during years 2 and 3, the FTP group both had higher income overall and derived a larger proportion of income from earnings.
[D] |
[D] |
- In the last few months of follow-up, the FTP group was less likely to receive welfare, but no more likely to work, and the two groups had about the same total income.
Table 3 summarizes FTP’s impacts in the last three months of the follow-up period.14 The results follow directly from the impact trends discussed above. Only 14 percent of the AFDC group was still receiving cash assistance by this point, but the receipt rate was only 8 percent for the FTP group. Interestingly, the difference — about 6 percentage points — is much smaller than the percentage of the FTP group that reached the time limit (17 percent). This suggests that many of the people who had their benefits canceled at the time limit would have left welfare anyway by the end of the follow-up period.
The reduction in average AFDC/TANF payments was very large in percentage terms — 48 percent — but small in dollar terms: The FTP group received $45 less in cash assistance, on average, during the three-month period.15 There was virtually no difference between the groups in average earnings, but the welfare reduction was so small that the two effects almost offset one another. As a result, combined income from AFDC/TANF, Food Stamps, and earnings was only slightly lower for the FTP group (the difference is not statistically significant). As noted earlier, the administrative records do not provide a full picture of household income.16 Indeed, results from the four-year client survey, discussed below, show that household income for both groups was substantially higher than the amounts shown in Table 3. Nevertheless, the survey confirms that there was no difference between the groups even when income was measured more completely.
The income distribution results in Table 3 suggest that FTP made some families worse off financially during the final three months — it reduced the number of people in the $1,501 to $3,000 income bracket and increased the number in the lower bracket. This result may be related to the fact that FTP slightly reduced the number of nonworking people who received both cash assistance and Food Stamps and increased the number who received Food Stamps only — a pattern consistent with nonworking people’s having their welfare grants canceled at the time limit.
- Most of the employed people in both research groups worked full time or close to full time in jobs that paid low wages and offered few fringe benefits.
| Outcome | FTP Group |
AFDC Group |
Difference (Impact) |
Percentage Change |
|---|---|---|---|---|
| Income amounts | ||||
| Average earnings ($) | 1,345 | 1,328 | 16 | 1.2 |
| Average AFDC/TANF payments ($) | 49 | 94 | -45*** | -48.1 |
| Average Food Stamp payments ($) | 228 | 251 | -23 | -9.1 |
| Average total income from earnings, AFDC/TANF, and Food Stamps ($)a | 1,622 | 1,674 | -52 | -3.1 |
| Income brackets (%) | ||||
| $0 | 35.7 | 33.8 | 1.9 | 5.7 |
| $1-$1,500 | 25.4 | 21.1 | 4.3*** | 20.3 |
| $1,501-$3,000 | 16.0 | 23.0 | -7.0*** | -30.4 |
| $3,001-$4,500 | 14.1 | 14.8 | -0.7 | -5.0 |
| $4,501 or more | 8.8 | 7.3 | 1.5 | 20.7 |
| 50% or more of income is derived from earnings (%) | 44.0 | 45.0 | -1.0 | -2.1 |
| Income sources | ||||
| Ever employed (%) | 48.0 | 49.7 | -1.7 | -3.4 |
| Ever received AFDC/TANF (%) | 8.1 | 14.0 | -6.0*** | -42.5 |
| Ever received Food Stamps (%) | 32.2 | 34.1 | -1.9 | -5.6 |
| Earnings without AFDC/TANF or Food Stamps | 31.1 | 31.1 | 0.1 | 0.2 |
| Earnings with AFDC/TANF or Food Stamps | 16.9 | 18.6 | -1.7 | -9.3 |
| No earnings and | ||||
| AFDC/TANF and Food Stamps | 5.3 | 8.4 | -3.1*** | -37.2 |
| Food Stamps only | 10.5 | 7.5 | 2.9*** | 38.6 |
| AFDC/TANF only | 0.5 | 0.5 | 0.0 | -2.5 |
| No AFDC/TANF or Food Stamps | 35.7 | 33.8 | 1.9 | 5.7 |
| Sample size | 1,405 | 1,410 | ||
| SOURCES: MDRC calculations from Florida Unemployment
Insurance (UI) earnings records, AFDC/TANF records, and Food Stamp
records. NOTES: Dollar averages include zero values for sample members who were not employed or were not receiving AFDC/TANF or Food Stamps. A two-tailed t-test was applied to differences between
the FTP and AFDC groups. Statistical significance levels are indicated
as *** = 1 percent; ** = 5 percent; * = 10 percent. |
FTP had little or no impact on the kinds of jobs sample members held at the end of the study period. Table 4 shows the characteristics of the current or most recent job held by FTP group members who responded to the four-year survey. About 80 percent of the employed people reported working at least 30 hours per week; half were working 40 or more hours. Hourly wages were generally low: Around three-fourths of respondents earned less than $7.50 per hour, and the overall average was about $6.90 per hour. Overall, 54 percent were working 30 or more hours per week in a job that paid less than $7.50 per hour.
Less than half of the employed people were in jobs that offered health insurance, and only about one-fourth were actually covered by employer health insurance (most of those who did not enroll in their company’s plan said it was too expensive or that they had not worked long enough to qualify for benefits).17 About one-third of the employed people in each group worked in jobs that provided paid sick days, a critical benefit for working parents. Finally, about one-third worked at night or had an irregular shift — schedules that can make it difficult to arrange stable child care arrangements.
- FTP had no impact on a range of measures of family structure and economic well-being although, on a few indicators, the FTP group’s living conditions appeared to be slightly better at the four-year point; levels of material hardship were high for both groups.
The four-year survey included information on household composition and income, family outcomes, and measures of economic well-being. As shown in Table 5, FTP slightly reduced the proportion of respondents who reported two or more housing problems (for example, roaches or broken windows) and four or more neighborhood problems (for example, drug users or pushers), and it increased the percentage who reported that, at the end of the month, they usually had enough money to make ends meet. In addition, FTP appears to have increased the percentage of families who received child support payments, an impact which could have been driven by programmatic efforts to enhance child support enforcement or by the need to replace welfare benefits lost at the time limit.18
At the same time, despite the modest income gains earlier in the follow-up period, FTP had no impact on overall material hardship, food security, health insurance coverage, vehicle ownership, or a range of other measures. FTP also did not affect fertility, marital status, or the composition of sample members’ households (interestingly, more than half the respondents in each group reported that they were living with at least one other adult when interviewed). Finally, as noted earlier, the survey confirms that household income was virtually the same for the two groups at the end of the study period.
| Characteristic | Outcome |
|---|---|
| Hourly wage (%)a | |
| Less than $6 | 42.6 |
| $6-$7.49 | 31.3 |
| $7.50-$8.99 | 9.6 |
| $9 or more | 16.6 |
| Average hourly wage ($) | 6.90 |
| Hours per week (%) | |
| Less than 20 | 4.7 |
| 20-29 | 15.6 |
| 30-39 | 28.0 |
| 40 or more | 51.7 |
| Average hours per week | 35.6 |
| Works at
least 30 hours per week in a job paying less than $7.50 per hour (%) |
54.3 |
| Job provides (%) | |
| Health insurance | 46.1 |
| Sick leave | 34.9 |
| Paid vacation | 45.0 |
| Respondent covered by employer health plan (%) | 26.9 |
| Work schedule (%) | |
| Day shift | 68.5 |
| Night shift | 17.0 |
| Irregular shift | 15.0 |
| Sample size | 787 |
| SOURCE: MDRC calculations from the four-year
client survey data. NOTES: The sample includes FTP group members who responded to the survey and who had ever worked since random assignment. aHourly wages are computed from other survey responses. Rounding may cause slight discrepancies in the calculation of sums and differences.(back) |
| Measure | FTP Group |
AFDC Group |
Difference (Impact) |
|---|---|---|---|
| Average number living in household | 3.9 | 3.9 | 0.0 |
| Average number of children in household | 2.1 | 2.2 | 0.0 |
| Respondent lives with at least one other adult (%) | 46.6 | 46.6 | 0.0 |
| Respondent gave birth since random assignment (%) | 23.9 | 22.7 | 1.2 |
| Respondent currently married and living with spouse (%) | 17.2 | 19.1 | -1.9 |
| Average household income in month prior to interview ($) | 1,469 | 1,379 | 89 |
| Respondent received child support in prior month (%) | 29.5 | 21.9 | 7.6 *** |
| Respondent owns a car, van, or truck (%) | 59.1 | 60.2 | -1.1 |
| Respondent has no health insurance (%) | 39.3 | 38.4 | 0.9 |
| Children have no health insurance (%) | 16.9 | 15.7 | 1.2 |
| Two or more housing problems (%)a | 14.1 | 18.4 | -4.3 ** |
| Four or more neighborhood problems (%)b | 17.2 | 21.0 | -3.8 * |
| Food insecure (%)c | 34.1 | 35.8 | -1.7 |
| Four or more material hardships (%)d | 18.3 | 19.9 | -1.7 |
| Two or more social services used (%)e | 19.2 | 19.2 | 0.0 |
| Usually has enough money at the end of the month (%) | 69.0 | 63.0 | 6.0 *** |
| Sample size | 860 | 869 |
| SOURCE: MDRC calculations from the four-year
client survey data. NOTES: A two-tailed
t-test was applied to differences between the FTP and AFDC groups.
Statistical significance levels are indicated as *** = 1 percent;
** = 5 percent; * = 10 percent. Rounding may cause slight discrepancies
in the calculation of sums and differences. |
Although FTP did not produce these conditions, the rates of material hardship were high for both groups: Nearly two-thirds of each group reported that they had experienced at least one serious material hardship in the past year — for example, being unable to pay their full rent or having their telephone disconnected.
- The employment and earnings gains were concentrated among less disadvantaged sample members; conversely, FTP had little or no impact on employment or earnings for more disadvantaged groups.
Often, overall results mask different patterns of impacts for particular subsets of people. Thus, the analysis examined FTP’s impacts separately for a variety of subgroups defined by characteristics that are associated with long-term welfare receipt and barriers to employment (for example, sample members’ employment and welfare histories before entering the study).
In general, these subgroup analyses found that FTP’s effects on employment and earnings were concentrated among less disadvantaged subgroups. For example, Table 6 summarizes FTP’s impacts for three subgroups: those most at risk of long-term welfare receipt (the right-hand column), those least at risk (the left-hand column), and those at medium risk (the middle column). Sample members were classified according to their employment and welfare history and other characteristics measured at the point they entered the study.
The top panel of the table, which displays results for the entire four-year follow-up period, shows that AFDC group members in the least at-risk subgroup had substantially higher earnings and substantially lower public assistance payments than their counterparts in the most at-risk group. Nevertheless, FTP increased earnings for the least at-risk subgroup by $4,221 (19 percent). In contrast, FTP generated no statistically significant earnings effects for the most at-risk subgroup. A similar pattern is evident in year 4, shown in the bottom panel.19
It is not clear why FTP was less effective at increasing employment and earnings for more disadvantaged participants. Most other studies of welfare-to-work programs have not found this pattern of results.20 Further analysis (not shown) found that a large proportion of these participants were placed into adult basic education while in FTP, and the disappointing results could be related to that particular activity. In addition, perhaps because of the strong local economy, it appears that the most disadvantaged members of the AFDC group had higher employment rates than similar individuals in other programs studied by MDRC over the past 15 years. The relatively strong AFDC group outcomes may have made it more difficult for FTP to generate significant impacts on employment-related outcomes.
Table 6 also shows that while FTP reduced cash assistance payments for all three subgroups, these reductions were smallest for the least at-risk group. This is not surprising, because relatively few people in this group would have been heavily dependent on welfare even without FTP (as illustrated by the AFDC group outcomes). Conversely, the reductions in cash assistance were fairly large — $1,087 (14 percent) over the four years and $518 (53 percent) in year 4 alone — for the most at-risk group, which was most likely to reach the time limit.
| Outcomes | Least at Risk of Long-Term Dependence | Medium Risk of Long-Term Dependence | Most at Risk of Long-Term Dependence | Sub- group Differ- ences |
||||||
|---|---|---|---|---|---|---|---|---|---|---|
| FTP Group | AFDC Group | Differ- ence |
FTP Group | AFDC Group | Differ- ence |
FTP Group | AFDC Group | Differ- ence |
||
| Entire follow-up period | ||||||||||
| Average total earnings ($) | 26,935 | 22,714 | 4,221*** | 13,888 | 11,867 | 2,021** | 12,048 | 10,571 | 1,477 | |
| Average total AFDC/TANF payments ($) | 1,726 | 2,216 | -490** | 3,647 | 4,311 | -664*** | 6,895 | 7,982 | -1,087 *** | |
| Average total Food Stamp payments ($) | 3,370 | 3,901 | -531** | 5,626 | 6,175 | -549** | 9,807 | 10,280 | -473 | |
| Average total income from earnings, AFDC/TANF, and Food Stamps ($)a | 32,031 | 28,831 | 3,200** | 23,160 | 22,353 | 807 | 28,750 | 28,832 | -82 | |
| Year 4 | ||||||||||
| Average quarterly employment rate (%) | 60.1 | 53.9 | 6.3* | 45.6 | 44.3 | 1.3 | 47.6 | 49.4 | -1.7 | |
| Average total earnings ($) | 7,760 | 6,613 | 1,147** | 4,414 | 4,013 | 402 | 4,219 | 3,930 | 288 | |
| Average quarterly AFDC/TANF receipt rate (%) | 5.8 | 9.2 | -3.4** | 12.1 | 20.2 | -8.2*** | 17.7 | 33.0 | -15.3 *** | *** |
| Average total AFDC/TANF payments ($) | 131 | 217 | -87** | 254 | 503 | -249*** | 451 | 969 | -518 *** | *** |
| Average Food Stamp receipt rate (%) | 22.4 | 24.0 | -1.5 | 37.8 | 39.4 | -1.6 | 61.8 | 60.1 | 1.7 | |
| Average total Food Stamp payments ($) | 494 | 504 | -11 | 926 | 1,032 | -106 | 1,978 | 1,928 | 50 | |
| Average total income from earnings, AFDC/TANF, and Food Stamps ($)a | 8,384 | 7,334 | 1,050* | 5,595 | 5,548 | 47 | 6,648 | 6,828 | -180 | |
| Sample size | 352 | 353 | 701 | 704 | 352 | 353 | ||||
| SOURCES: MDRC calculations from Florida Unemployment
Insurance (UI) earnings records, AFDC/TANF records, and Food Stamp
records. NOTES: A two-tailed t-test was applied to differences between the FTP and AFDC groups. Statistical significance levels are indicated as *** = 1 percent; ** = 5 percent; * = 10 percent. An F-test was performed to determine whether the
variation in impacts across subgroups was statistically significant.
These results are presented in the final column of the table. Statistical
significance levels are indicated as *** = 1 percent; ** = 5 percent;
* = 10 percent. |
The combined effect of the earnings and cash assistance results was that FTP substantially raised total income for the least at-risk group, both over the full period and in year 4 alone — their earnings gains far outweighed their losses in public assistance. In contrast, for the most at-risk group, the welfare reductions offset the small (statistically insignificant) earnings gains, resulting in no impact on total income.
Further analysis (not shown) found that, for a small subset of the most at-risk group facing particularly serious barriers to employment (long-term welfare recipients with no high school diploma and no recent work history), the FTP group had about $2,000 less combined income than the AFDC group over the four-year period. This subgroup experienced even smaller earnings gains, and larger welfare reductions, than the full most at-risk group shown in Table 6. This result should be interpreted with caution, however, because the income loss, while large in dollar terms, is not statistically significant. Also, there is little evidence that the loss translated into increases in material hardship or changes in household composition measured via the four-year client survey. It is possible that FTP group households within the subgroup had more income from sources not measured in the administrative records (data are not available to examine this issue).21
D. FTP’s Impacts on Outcomes for Families and Children
The four-year client survey asked parents a small number of questions about recent child care arrangements, school outcomes, and delinquent behavior for each of their children. In addition, respondents who had at least one child between 5 and 12 years old at the time of the survey answered a set of detailed questions about child care use, father’s involvement, parenting, school performance, and other outcomes for one “focal” child in that age range.22 Key findings include:
- FTP children spent more time in child care than their AFDC group peers, and they were more likely to have contact with their noncustodial fathers.
Table 7 shows the current child care arrangements for all children under 5 years old at the point the four-year survey was administered, as well as for those between 5 and 12. The table shows that FTP increased the percentage of children in child care for both age groups (although not shown in the table, FTP did not increase child care among children over 12). The table also shows that most children were being cared for by relatives or other informal providers, rather than in child care centers or pre-schools. Among the children under age 5, FTP increased the already sizable proportion who were in care more than 20 hours per week. A more detailed analysis of the 5- to 12-year-old focal children (not shown) found that the increase in child care was not accompanied by an increase in the number of children in unstable child care arrangements or in low-quality child care settings (as perceived by parents). Analyses of administrative data (also not shown) found that child care subsidies were more likely to be provided for children in the FTP group relative to those in the AFDC group, although there were no differences between the two groups by the fourth year of follow-up.
| Outcome | Ages 0-4 | Ages 5-12 | ||||
|---|---|---|---|---|---|---|
| FTP Group |
AFDC Group |
Difference (Impact) |
FTP Group |
AFDC Group |
Difference (Impact) |
|
| Currently in child care | 48.1 | 41.2 | 6.9 * | 39.6 | 35.2 | 4.4 ** |
| Relative care (%) | 26.3 | 23.6 | 2.7 | 26.2 | 23.1 | 3.1 * |
| Nonrelative care (%) | 9.0 | 6.5 | 2.5 | 5.3 | 5.2 | 0.0 |
| Formal care (%) a) | 14.1 | 13.3 | 0.8 | 11.3 | 9.6 | 1.7 |
| Hours in child care in a typical week | ||||||
| Less than 20 (%) | 7.7 | 9.3 | -1.6 | 20.9 | 16.5 | 4.4 *** |
| 20 or more (%) | 39.2 | 31.7 | 7.5 ** | 17.8 | 18.2 | -0.4 |
| Sample size (total = 1,877) | 331 | 325 | 1,125 | 1,176 | ||
| SOURCE: MDRC calculations from the four-year
client survey. NOTES: A two-tailed t-test
was applied to regression-adjusted impact estimates. Statistical
significance levels are indicated as ***=1 percent; **=5 percent;
*=10 percent. Rounding may cause slight discrepancies in the calculation
of sums and differences. |
Although not shown in the table, FTP also increased the percentage of 5- to 12-year-old focal children who had been cared for by their noncustodial father in the past year. As noted earlier, it also increased financial contributions from noncustodial fathers. However, it is important to note that overall rates of father involvement were relatively low. For example, less than 30 percent of FTP group focal children with a living noncustodial father saw their father at least monthly, and more than 40 percent had not seen their father at all in the past year.
- Overall, FTP had few effects across a range of measures of parenting and child well-being for 5- to 12-year-olds; there were a couple of negative impacts on school-related outcomes for adolescents, however.
As shown in the top panel of Table 8, there were few significant differences between FTP and AFDC group focal children on school, behavior, and health measures, and those that were significant did not consistently favor one group or the other. Also, parents in the two groups did not differ on most measures of their emotional health or parenting behavior (not shown in the table).
In contrast to the results for 5- to 12-year-olds, FTP had a couple of negative impacts for adolescent children (ages 13 to 17): As shown in the bottom panel of Table 8, 41 percent of FTP group adolescents had been suspended from school at least once since random assignment (compared with 33 percent of AFDC group adolescents), and average school performance (as reported by parents) was somewhat lower for the FTP group. However, there were no differences between groups on a number of other measures of school performance and behavior.
- Surprisingly, FTP generated some negative effects for children in the least disadvantaged families — the subgroup with the largest earnings impacts.
Table 9 shows FTP’s impacts on several school-related measures for school-age children in the three subgroups discussed earlier. As the table shows, FTP had negative effects on school achievement and increased school suspensions for children in the families who were least at risk of long-term welfare dependence. A more detailed analysis of the 5- to 12-year old focal children (based on a small sample) found that FTP parents in the least at-risk subgroup supervised their children less closely than did AFDC group parents, perhaps because they were more likely to be working near the end of the follow-up period and their children had worse outcomes on behavioral and school measures. Interestingly, unfavorable impacts were generally not found for the medium-risk group; this group experienced employment impacts earlier in the follow-up period, but these impacts faded during year 4.
| Outcome | FTP Group |
AFDC Group |
Difference (Impact) |
Percentage Change |
|---|---|---|---|---|
| Focal children ages 5-12 | ||||
| School outcomes | ||||
| Average achievementa | 4.1 | 4.0 | 0.1 | 2.5 |
| Below average (%) | 7.4 | 9.5 | -2.1 | -22.3 |
| Since random assignment, child | ||||
| Ever in special education (%) | 12.3 | 10.1 | 2.2 | 21.9 |
| Ever suspended (%) | 8.2 | 8.8 | -0.6 | -6.5 |
| Behavior | ||||
| Behavioral Problems Index total scoreb | 10.8 | 10.9 | -0.1 | -0.7 |
| Positive Behavior Scale total scorec | 59.0 | 60.2 | -1.2 * | -2.0 |
| Health | ||||
| General healthd | 4.2 | 4.1 | 0.1 * | 2.2 |
| Sample size (total = 1,108) | 543 | 565 | ||
| Adolescents ages 13-17 | ||||
| School outcomes | ||||
| Average achievementa | 3.7 | 3.9 | -0.2 * | -4.0 |
| Below average (%) | 14.8 | 10.9 | 3.9 | 36.0 |
| Since random assignment, child | ||||
| Ever in special education (%) | 18.7 | 15.4 | 3.3 | 21.7 |
| Ever suspended (%) | 40.7 | 32.7 | 8.0 ** | 24.4 |
| Behavior | ||||
| Child ever arrested (%) | 9.6 | 9.2 | 0.4 | 4.1 |
| Child ever had a baby (%) | 2.8 | 3.3 | -0.5 | -16.1 |
| Sample size (total = 741) | 367 | 374 | ||
| SOURCE: MDRC calculations from the four-year
client survey. NOTES: A two-tailed t-test was applied to regression-adjusted impact estimates. Statistical significance levels are indicated as *** = 1 percent; ** = 5 percent; * = 10 percent. Rounding may cause slight discrepancies in the calculation of sums and differences. aMothers
were asked to rate their child's overall perfomance in school from
1 (doing "not well at all") to 5 (doing "very well").(back)
|
| Outcome | Least at Risk of Long-Term Dependence | Medium Risk of Long-Term Dependence | Most at Risk of Long-Term Dependence | Sub- group Differ- ences |
||||||
|---|---|---|---|---|---|---|---|---|---|---|
| FTP Group | AFDC Group | Differ- ence (Impact) |
FTP Group | AFDC Group | Differ- ence (Impact) |
FTP Group | AFDC Group | Differ- ence (Impact) |
||
| Average achievementa | 3.9 | 4.2 | -0.3*** | 4.0 | 4.0 | 0.1 | 3.8 | 3.8 | 0.1 | *** |
| Below average (%) | 13.7 | 7.3 | 6.4** | 8.9 | 8.7 | 0.3 | 10.1 | 13.1 | -3.0 | ** |
| Since random
assignment, child Ever in special education (%) |
15.3 | 13.1 | 2.2 | 12.8 | 9.9 | 2.9 | 13.9 | 14.5 | -0.5 | |
| Sample size (total= 3,042) | 276 | 293 | 693 | 690 | 523 | 567 | ||||
| Ever suspended (ages 10 and older) (%) | 34.3 | 22.0 | 12.3** | 27.3 | 28.2 | -0.9 | 27.7 | 26.7 | 1.0 | * |
| Ever expelled (ages 10 and older) (%) | 5.1 | 2.1 | 3.0 | 5.7 | 2.5 | 3.2 ** | 1.8 | 3.8 | -2.1 | ** |
| Sample size (total= 1,425) | 167 | 177 | 315 | 313 | 218 | 235 | ||||
| SOURCE: MDRC calculations from the four-year
client survey. NOTES: The sample includes
families with children ages 5-17 at the time of the four-year interview
who were randomly assigned from August 1994 to February 1995. A
two-tailed t-test was applied to regression-adjusted impact estimates.
Statistical significance levels are indicated as *** = 1 percent;
** = 5 percent; * = 10 percent. Standard errors were adjusted to
account for shared variance between siblings. |
Notably, FTP had little or no impact on children in the most disadvantaged families, whether defined as those at highest risk of long-term dependence (shown in the table) or the subset of that group facing multiple barriers to employment (not shown).
E. After the Time Limit
MDRC used a variety of data sources to examine the post-welfare experiences of the 237 report sample members who reached the time limit: All were tracked using administrative records, and some responded to the four-year survey. In addition, as part of a special study, 54 were interviewed in depth around the time their benefits expired and then 6, 12, and 18 months later. These interviews provide rich descriptive information but cannot be used to assess the impact of the time limit because there is no way to know for sure what would have happened to these 237 people had they been allowed to remain on welfare.23
- The post-welfare experiences of families whose grants were canceled varied considerably; most struggled financially, but did not appear to be worse off than many other families who left welfare for other reasons.
According to administrative records, just over 40 percent of those who were terminated from welfare worked in all four quarters of the subsequent year (these results are not shown in a table). On the other hand, 36 percent worked in none or only one of the quarters. The overall employment rate for the individuals who reached the time limit was about the same in the year after the time limit as it was in the year before. However, average earnings were substantially higher after the time limit, suggesting that some of these individuals worked more often after their benefits were cut off.
The in-depth interviews found that most of those who worked sporadically or not at all in the post-time-limit period relied heavily on a parent, partner, or spouse. Many lived in homes belonging to family members and paid little or no rent (in many cases, these living arrangements began long before the family reached the time limit) or in public or subsidized housing, where their rent was pegged to their income. The vast majority received Food Stamps. Several respondents chose not to work because they wanted to care for their children or continue their education. A few wanted to work but could not find (or hold) jobs; they were surviving on a limited and precarious mix of Food Stamps, housing assistance, and irregular income sources.
Overall, instances of extreme material hardship such as homelessness and hunger were quite rare, but almost all the families struggled financially (as they had before reaching the time limit). Interestingly, levels of material hardship were not strongly correlated with employment status. In fact, on some measures, the working families — who tended to receive less support from family members and from public assistance — appeared to be experiencing greater levels of hardship than the nonworking families. But it is impossible to trace the direction of causality: Were the nonworking people not working because they couldn’t work or because they had other supports that allowed them not to work? And, conversely, were the working families working because they had fewer other supports, or did they need less help because they were working?
Finally, responses to the four-year client survey indicate that the families whose grants were terminated at the time limit did not appear to be experiencing greater levels of material hardship than other FTP (or AFDC) families who left welfare for other reasons. A key question is whether this will continue to be the case over time, because the terminated families have lost access to the cash assistance safety net.
F. Financial Costs and Benefits of FTP
- Owing to its enhanced services and supports, FTP cost about three times as much, per person, as traditional AFDC combined with Project Independence.
As a relatively small pilot program, designed at a point when welfare time limits were not widely accepted, FTP was quite generously funded. Florida approached time limits cautiously, embedding the limit in a program that was very heavily staffed and that offered an unusually rich array of services and supports. Not surprisingly, costs were high: FTP’s five-year net cost — the per person cost of FTP above and beyond what would have been spent under AFDC and Project Independence — was nearly $8,000 per person, a figure at the high end of programs evaluated by MDRC (the gross costs of FTP and AFDC/PI were about $12,500 and $4,500 per person, respectively).
About 40 percent of the increased cost was attributable to FTP’s enhanced employment-related services — the services themselves (and the associated staffing) were more expensive than traditional PI services, and, as noted earlier, the rates and levels of participation in these services were much higher under FTP. The higher levels of participation in these activities, along with higher rates of employment and more generous funding in FTP, also generated much higher costs for child care, transportation, and other support services; these accounted for another 30 percent of FTP’s net cost. The remaining component of the net cost was mostly attributable to the very small caseloads of FTP case managers.
- From the government budget perspective, the public assistance savings generated by FTP were not large enough to offset its costs; FTP participants, however, experienced a small financial gain, on average.
As noted earlier, FTP’s ability to generate budgetary savings by reducing cash assistance receipt was limited by the fact that the AFDC group left welfare so rapidly. Thus, savings for taxpayers did not come close to offsetting the program’s net costs, although saving money was never emphasized as a key program goal. In addition, there is no way to know whether the program would have achieved its impacts on earnings or other outcomes if staffing and service levels had been lower.
As might be expected given the income data reported earlier, FTP participants benefited financially: Projected over a five-year period, their higher earnings (supplemented by the federal Earned Income Credit) outweighed their income losses (lower public assistance benefits, higher payroll taxes, etc.) by a little over $1,500 per person, on average.
III. Policy Implications
The FTP evaluation provides some of the first information on the implementation and impacts of a welfare reform strategy that included a time limit on benefit receipt. Judged against its own goals — which focused heavily on reducing dependency — FTP was relatively successful. It substantially reduced long-term welfare receipt and, at least during the study period, did not produce the very harmful impacts some people had predicted. Unlike some other welfare-to-work models, FTP did not save money for taxpayers, but that was not an explicit goal; in part, the state used the relatively small pilot to learn more about what level of resources would be needed for a program of this type. Similarly, FTP’s impacts on family income and other measures of economic well-being were both smaller and less sustained than those generated by other models that were explicitly designed both to raise earnings and to reduce poverty.24
The results provide some lessons on other issues relevant to the current environment:
The impact of benefit termination. Because FTP was the first program in which families were cut off welfare at a time limit, the evaluation provides one of the first opportunities to examine a central question raised by the welfare reforms of the 1990s: How will families fare after they are terminated from cash assistance?25
Unfortunately, it turns out that this question is extraordinarily difficult to answer in a rigorous way. It is fairly clear that the most extreme claims of both advocates and critics of time limits have not come to pass in Escambia County. MDRC’s in-depth examination of the terminated families over an 18-month period uncovered few dramatic success stories, but equally few instances of extreme deprivation. Of course, the situation may change — for better of worse — over a longer follow-up period.26
But were the families better off or worse off? From a simple before-and-after perspective, they obviously lost income when their welfare checks were canceled. It appears that some of them had managed to replace the lost income 18 months later, while others had not (although their situations were extremely fluid).
But the real question is: Are the terminated families better off or worse off than they would have been had FTP not existed? Here, the answer is much more complicated. For example, it is clear that some of the terminated families were initially better off than they would have been because they went to work before reaching the time limit and FTP’s enhanced earnings disregard allowed them to supplement their earnings with a partial welfare grant. When they were cut off, they were brought back to where they would have been without the disregard (although without the option of returning to welfare later). In addition, the impact results show that many of those who were terminated at the time limit would have left welfare anyway shortly thereafter. In contrast, other FTP participants were terminated without jobs and would have remained on welfare had it not been for FTP; it seems likely that these families were made worse off financially, although perhaps not dramatically so because of Florida’s low grant levels.
In any case, in drawing conclusions from these results, it is critical to reiterate that FTP did not terminate all families who received 24 or 36 months of benefits. The program cut off nearly all of those who actually reached the time limit, but a significant number of participants were granted exemptions that stopped their time-limit clocks (or they were exempted before their clock started); in a few other cases, the children’s portion of the grant was retained. These families might have experienced more serious problems had their grants been closed. Similarly, as noted earlier, the consequences might have been quite different in a larger city, a weaker labor market, or a state with higher benefit levels.
Earnings disregards and time limits. Like Florida, most states have chosen to impose time limits and simultaneously expand earnings disregards (although the enhanced disregard was not a main focus of FTP). Studies have shown that earnings disregards, when combined with employment-related mandates, can raise employment and income, and FTP’s disregard is at least partly responsible for the income gains generated by the program. Nevertheless, the enhanced disregard also caused some families to use up their months of benefits faster than they otherwise would have. Moreover, combining these policies complicates the program message: It is difficult to urge recipients both to leave welfare quickly in order to “bank” their available months and to take advantage of a disregard by combining work and welfare.
One way to make the message more consistent is to stop the time-limit clock for recipients who are working and receiving welfare. Illinois, Rhode Island, and a handful of other states have done this. In effect, their time limits apply to welfare without work. This strategy implicitly assumes that some families should receive longer-term income supplementation, given the prevalence of low-wage jobs.
Implementing time limits. One of the critical questions in implementing time limits is how to decide which families should qualify for safeguards such as exemptions or extensions. FTP chose not to create explicit definitions of key terms such as “compliant” but implemented a detailed, multistage review of each case. The impact results suggest that this process succeeded in identifying and protecting (via exemptions or partial terminations) some of the participants facing very serious problems. But FTP’s labor-intensive process might not be replicable in a larger program, and, without such a process, the lack of explicit guidelines might make it difficult to ensure that all recipients receive equal treatment.
Effects on children. FTP had few impacts on child well-being overall, but the impacts that occurred were somewhat unexpected. Many observers have warned that pushing single mothers into the labor force might produce negative impacts on young children, who would be forced to spend more time in low-quality child care arrangements. Although FTP increased the amount of time children spent in child care, it did not appear to increase time in unsafe or unstimulating care. There were also no impacts on school-related outcomes for children who were 1 to 8 when their parents entered the program. On the other hand, FTP adolescents appear to have performed somewhat worse than their AFDC group counterparts on selected measures. This result is consistent with another recent study, suggesting that increases in maternal employment may have negative consequences for certain groups of older children.27
Similarly, some predicted that children in the most disadvantaged families were most at risk of harm. In fact, FTP’s negative impacts for children were concentrated among the least disadvantaged families, the group least likely to be directly affected by the time limit (but with the largest earnings gains). Of course, the pattern might have been different for the most disadvantaged if the time limit had been implemented in a different way (for example, if no exemptions had been granted).
Supports for working families. Four years after enrollment, most FTP families were still struggling. Most were working, but few had moved out of poverty. A large fraction had no health insurance, and food insecurity and other material hardships were prevalent. These outcomes were not caused by FTP — on average, the program had little or no impact in any of these areas. In addition, given Florida’s low grant levels, most of these families were probably better off financially than a family surviving on only cash assistance and Food Stamps. Nevertheless, the outcome levels for both groups highlight the importance of additional supports for low-income working families, particularly if such families will be expected to stay off welfare for long periods.
1FTP was initially implemented in two counties, Escambia (discussed in this report) and Alachua, which operated a version of FTP in which participation was voluntary. MDRC produced a single report on the impacts of the Alachua program before it was phased out in 1996. Several other counties briefly implemented FTP in 1996; they are not part of the study. (back)
2Recipients were limited to 36 months of welfare in any 72-month period if they (1) had received AFDC for at least 36 of the 60 months prior to enrollment or (2) were under 24 years old and had no high school diploma and no recent work experience. (back)
3Most of these states have imposed “reduction” time limits that eliminate the adult’s portion of a family’s welfare grant but leave the children’s portion intact. Two states have imposed no time limit. If these policies remain in place, all eight of these states will need to use state funds to assist children or entire families who pass the federal 60-month limit and exceed the cap on exemptions. (back)
4Both FTP and WAGES set time limits of 24 months in any 60-month period for most recipients and 36 months in any 72-month period for the least job-ready. However, unlike FTP, WAGES also imposed a 48-month lifetime time limit on benefit receipt. (back)
5The following groups were exempt from FTP; they were screened out and not randomly assigned: “child-only” cases in which no adult was counted in the grant calculation; recipients who were incapacitated or caring full time for a disabled dependent; recipients who were under 18 and in school or working; recipients who were 62 years old or older; and parents caring for a child under 6 months old. A narrower range of families was exempted under WAGES. (back)
6The four-year client survey targeted a subset of the report sample — the 2,160 people randomly assigned from August 1994 to February 1995. (back)
7Although not shown in the table, the percentage who strongly agreed with the statement was identical for the two groups — 33 percent. (back)
8Individuals who gave birth after entering FTP were exempt from mandatory participation in employment-related activities until their child was 7 months old, but their time-limit clock continued to run. (back)
9Florida officially canceled its waiver after the 1996 federal welfare law passed, but it continued to operate FTP according to the waiver’s terms and conditions in order to avoid disrupting the evaluation. (back)
10The numbers in Figure 3 do not precisely match those cited in the previous section. For example, Figure 3 shows that 18 percent of those subject to a 24-month time limit accumulated 24 months of benefits (139/768), while the earlier section says this figure is 16 percent. The difference is that the earlier section measured benefit receipt within four years after random assignment for each person. Figure 3 follows each person through June 1999, a follow-up period of 52 to 61 months (depending on the individual’s random assignment date). (back)
11The federal waiver required that the transitional jobs would allow former recipients to earn at least as much as the standard AFDC grant for their family size, plus a $90 allowance for work expenses. This became FTP’s definition of self-sufficiency because families with at least this much income from non-welfare sources would presumably be no worse off after leaving welfare than they would have been had they been receiving welfare and not working. Officially, the requirement to provide transitional jobs also applied to people who were earning grant plus $90 at the time limit but later became unemployed, but FTP did not implement this provision. (back)
12Factoring in the Earned Income Credit, however, does not change the impact on income. Although it is estimated that the FTP group received nearly $300 more than the AFDC group from this credit over the four-year period, that increase was offset by increased taxes the FTP group paid. (back)
13Nonexperimental analysis using data from the FTP study support this hypothesis. See Jeffrey Grogger and Charles Michalopoulos, “Welfare Dynamics Under Time Limits,” NBER Working Paper No. W7353, September 1999. (back)
14These results are for the second quarter of year 5, slightly beyond the period summarized in Table 2. (back)
15All of the dollar amounts in the table are averages that include zero values for those who did not work or receive welfare during the period. FTP group members who received AFDC/TANF received $605 during the quarter, on average. Those who worked earned an average of $2,802. (back)
16Table 3 shows that more than one-third of each group had no income from UI-covered earnings, cash assistance, or Food Stamps in the last three months of follow-up. Further analysis using survey data (not shown in the table) found that almost all of these sample members had income from other sources (for example, child support or non-UI earnings) and/or were living with other adults who had income. (back)
17Of those who were offered employer health insurance but did not enroll, about half reported that they were covered by Medicaid or some other insurance; the rest were uninsured. (back)
18In part, the impact on child support receipt may have occurred because AFDC group members were more likely to be on welfare when interviewed, and thus less likely to be aware that child support was being collected on their behalf (child support collected for children on welfare is mostly retained by the state as reimbursement for welfare costs). However, the fact that FTP also increased the proportion of children who had been cared for by their noncustodial fathers (see below) lends some additional credibility to the child support impact. (back)
19This pattern of subgroup results should be interpreted cautiously because the differences in earnings impacts between groups are not statistically significant. (back)
20See Charles Michalopoulos and Christine Schwartz, What Works Best for Whom: Impacts of 20 Welfare-to-Work Programs by Subgroup, National Evaluation of Welfare-to-Work Strategies (Washington, DC: U.S. Department of Health and Human Services, Administration for Children and Families and Office of the Assistant Secretary for Planning and Evaluation; and U.S. Department of Education, Office of the Under Secretary and Office of Vocational and Adult Education, 2000). (back)
21The four-year client survey provides information on all sources of household income, but only for the month prior to the interview. For the most part, the income losses measured with administrative records occurred earlier in the follow-up period. (back)
22The focal children were chosen before the survey was administered by identifying all single mothers who had a child between 1 and 8 years of age at the point of random assignment (these children were between 5 and 12 four years later). When a sample member had more than one child in the age range, one was chosen at random as the focal child. (back)
23In general, the AFDC group provides a benchmark for assessing outcomes for the FTP group, but it is difficult to determine which subset of the AFDC group would serve as the most appropriate benchmark for assessing the experiences of the FTP participants who reached the time limit. (back)
24See, for example, Cynthia Miller et al., Reforming Welfare and Rewarding Work: Final Report on the Minne-sota Family Investment Program, Vol. 1, Effects on Adults (New York: MDRC, 2000). (back)
25Of course, some of the individuals who were affected by the time limit never reached it; they were motivated to find jobs and leave welfare before accumulating 24 or 36 months of receipt. (back)
26It is difficult to predict what might happen: Owing to the design of FTP’s time limit, the terminated families will eventually be allowed to return to welfare. (back)
27Pamela Morris and Charles Michalopoulos, The Self-Sufficiency Project at 36 Months: Effects on Children of a Program That Increased Parental Employment and Income (Ottawa: Social Research and Demonstration Corporation, 2000). (back)
| Table of Contents | Previous | Next |