[ Main page of Report | Contents of Report ]
[ Go to Contents ]
We utilize the random assignment design of the three evaluation sites to ask whether the program implemented at each site affected the likelihood of cycling. Each site's program was implemented pre-PRWORA and was designed to evaluate several aspects of change in welfare benefit receipt. While many of these policies were later enacted under PRWORA, they should not be seen as representing final PRWORA implementation. Our analysis is based on program group and control group differences in cycling rates, after controlling for the factors shown to affect cycling in Table 6.
The results of the analysis are presented in Table 12. The top panel of the table shows impacts for the full sample, which includes both new recipients to welfare and recipients in the middle of a welfare spell at the time of sample intake, whereas the lower panel shows impacts for new recipients.(38) This panel shows how results would differ when using a sample for which all spell lengths are known, at least since their month of the sample intake. The top panel of the table shows that the Connecticut Jobs First program had a small but statistically significant effect on the likelihood of cycling for the full sample. The program decreased the occurrence of welfare cycling by 2.9 percentage points, while increasing the likelihood of short-term recipiency by 8.2 percentage points.
The results for Florida FTP are similar in direction. Although the program did not affect the likelihood of welfare cycling, it did increase the likelihood of short-term recipiency and decreased the likelihood of long-term recipiency. In contrast, Vermont WRP appears to have had no effect on the likelihood of cycling or not cycling.
The lower panel of the table presents program impacts for the subsample of new recipients. The lower panel shows that the program effects largely disappear for new recipients, indicating that the effects were concentrated among ongoing recipients in Connecticut Jobs First and Florida FTP. The single exception is the program effect of Florida FTP on short-term recipiency, which appears to have increased the number of new recipients that became short-term recipients within the four-year observation period.
| Program (%) | Program Group | Control Group | Difference (Impact) |
|---|---|---|---|
| Full Sample | |||
| Connecticut Jobs First | |||
| Cyclers | 4.6 | 7.4 | -2.9 ** |
| Short-term recipients | 52.3 | 44.1 | 8.2 ** |
| Long-term recipients | 43.1 | 48.4 | -5.3 ** |
| Sample size | 2,184 | 2,102 | |
| Florida FTP | |||
| Cyclers | 13.6 | 14.8 | -1.2 |
| Short-term recipients | 59.1 | 52.0 | 7.1 ** |
| Long-term recipients | 27.3 | 33.2 | -5.9 ** |
| Sample size | 1,150 | 1,124 | |
| Vermont WRP | |||
| Cyclers | 9.4 | 8.0 | 1.4 |
| Short-term recipients | 42.2 | 41.9 | 0.3 |
| Long-term recipients | 48.4 | 50.1 | -1.7 |
| Sample size | 4,051 | 1,004 | |
| New Recipients | |||
| Connecticut Jobs First | |||
| Cyclers | 6.2 | 7.7 | -1.5 |
| Short-term recipients | 60.5 | 58.1 | 2.4 |
| Long-term recipients | 33.3 | 34.1 | -0.9 |
| Sample size | 860 | 858 | |
| Florida FTP | |||
| Cyclers | 11.6 | 14.8 | -3.2 |
| Short-term recipients | 75.8 | 68.4 | 7.5 ** |
| Long-term recipients | 12.6 | 16.9 | -4.3 |
| Sample size | 420 | 362 | |
| Vermont WRP | |||
| Cyclers | 11.4 | 9.8 | 1.6 |
| Short-term recipients | 56.3 | 56.2 | 0.1 |
| Long-term recipients | 32.3 | 34.0 | -1.7 |
| Sample size | 1,383 | 315 | |
| Sources: MDRC calculations from
state and county administrative records. Notes: Estimates were regression-adjusted using ordinary least squares, controlling for pre-random assignment (sample intake) characteristics of sample members and for environmental conditions. "**" indicates statistical significance at the 0.05 level or smaller. Results from logistic regressions performed on binary outcomes were consistent with the above results. |
|||
In summary, welfare policies that anticipated PWRORA — as represented by the Connecticut Jobs First, Florida FTP, and Vermont WRP programs — decreased cycling somewhat among ongoing recipients, but did not affect cycling among new recipients. This suggests that the programs were more effective on long-term recipients. This difference in findings between ongoing recipients and new recipients will be useful in interpreting the results in the next section, which analyzes the change in the incidence of cycling after welfare reform among a large sample of new recipients.
[ Go to Contents ]
Welfare caseloads have declined dramatically during the 1990s, especially during the years following passage of PRWORA. Undoubtedly, the changes to the welfare system mandated by PRWORA contributed to the nation-wide decrease in welfare receipt. However, the reduction in the welfare rolls began before August 1996, when PRWORA was enacted. Moreover, there are several other possible explanations for the decrease in caseloads. For instance, the growing economy during the mid- to late-90s possibly made it easier for current recipients and those at risk to find relatively high-paying jobs. The expansion of the federal Earned Income Tax Credit (EITC) may have encouraged many to leave welfare for work or to go to work instead of applying for welfare (Meyer and Rosenbaum, forthcoming). The aging of the population may have resulted in fewer families with children under the age of 18. Further, declines in out-of-wedlock childbearing may have resulted in fewer families being eligible for cash assistance (Sawhill, 2001).
These factors present difficulties for understanding the effects of PRWORA on cycling. All would be expected to possibly increase cycling, and all were present both before and after PRWORA was implemented.(39)
This section begins by looking descriptively at what happened to the welfare caseload and welfare exits over time by recipient type. Figure 3 shows the number of open welfare cases each month by recipient type in Cleveland for members of the special sample for studying the effects of PRWORA. As the figure shows, the proportion of active cases that cycle increased throughout the period. In December 1999, cyclers represented 18 percent of all active recipients, up from 10 percent in January 1993. During the pre-reform period (1993 to 1995) cyclers represented slightly more than 10 percent of all active recipients, while during the post-reform period (1997 to 1999) cyclers represented almost 18 percent of the caseload in Cleveland. In general, cycling increased since the passage of PRWORA, although is not clear how much of this increase is actually due to welfare reform.
Figure 3 also shows that the proportion of active long-term recipients peaked in 1996 and declined thereafter. By definition, this change in the proportion of long-term recipients was accompanied by a decrease and then an increase in the proportion of short-term recipients. In fact, in June 2000, short-term recipients became the dominant type of active recipient among sample members.
The two vertical lines in Figure 3 represent the periods when welfare reform may have first affected behavior in Cleveland. The first vertical line represents August 1996, when PRWORA was signed into federal law. The second vertical line represents the implementation of Ohio Works First, Ohio's TANF program. Note that the increase in the proportion of active cyclers and the decrease in the proportion of active long-term recipients appear to coincide with these reform dates.
Figure 5 shows similar information for Philadelphia during each month from January 1993 through December 2001. Compared with Cleveland, Philadelphia had a smaller proportion of cyclers throughout the time frame. In fact, by the end of the period less than 10 percent of active recipients were cyclers. However, there was an increase in cycling over time similar to Cleveland's. For example, during the pre-reform period (1993 to 1995), slightly more than 3 percent of active recipients were cyclers, compared with almost 7 percent over the post-reform period (1997 to 2001).
In Philadelphia, the percentage of active long-term recipients increased early in the period and then stabilized. That is, the trend in Figure 5 represents people who receive benefits each month who are, or eventually will be, long-term recipients. Unlike Cleveland, there was no apparent decrease in the percentage of active long-term recipients. In fact, the percentage of active short-term recipients in Philadelphia decreased over time. By the end of the period, there were only slightly more active short-term recipients than cyclers.
Sources: MDRC calculations from Cleveland administrative records.
Notes: The sample includes only new recipients during their month
of sample intakes. See Table 2 for sample intake
period for each site.
Sources: MDRC calculations from Cleveland administrative records.
Notes: The sample includes only new recipients during their month
of sample intakes. See Table 2 for sample intake
period for each site.
Sources: MDRC calculations from Cleveland administrative records.
Notes: The sample includes only new recipients during their month
of sample intakes. See Table 2 for sample intake
period for each site.
These descriptive results suggest that cycling was more prevalent in Cleveland than in Philadelphia at any particular point, although the incidence of cycling increased over time in both sites. While informative, these figures show all active recipients at a point in time and do not allow a clear look at trends in behavior as an entry cohort analysis, a variation of which is performed in the next section.
To understand how cycling behavior changes over time, consider Figure 4. Each point on the figure represents an outcome for a group of new recipients. For example, the left-most point on each line represents the outcomes for the first cohort of new recipients, who began receiving welfare in January 1993. The corresponding points at the far right of the diagram represent outcomes for the last cohort of new recipients, who began receiving welfare in December 1996.(40) The vertical line represents August 1996, the official signing of PRWORA. Points to the right of the vertical line represent outcomes for people who first began receiving welfare after PRWORA was enacted, while points to the left of the vertical line represent outcomes for people who began receiving welfare before PRWORA was enacted.
The outcome of greatest interest represented in Figure 4 is the percentage of new recipients in Cleveland who became cyclers (versus becoming either short-term or long-term recipients) within four years of first receiving benefits. This result is shown by the solid line closest to the x-axis. The left-most point on this line indicates that about 10 percent of adults who began receiving welfare in Cleveland in January 1993 became welfare cyclers by December 1996, four years later. The figure consequently provides information on whether the proportion of welfare cyclers changed over time and whether that change seemed related at all to the PRWORA reforms.
Figure 4 shows a fairly stable trend: later groups were slightly more likely to become welfare cyclers than earlier groups. For example, the right-most point indicates that more than 17 percent of people who began receiving welfare in December 1996 had become cyclers within four years (that is, by November 2000), compared with the 10 percent of the January 1993 group mentioned above.
A similar trend for Philadelphia is shown in Figure 6, although this trend is less pronounced than that of Figure 4. The left-most point of Figure 6 indicates that 4 percent of people who began receiving welfare in January 1993 went on to become cyclers by December 1996, four years later. By the end of the period, the number of people who became cyclers increased slightly. The right-most point of Figure 6 indicates that close to 9 percent of people who started receiving welfare for the first time in December 1996 went on to become cyclers by November 2000. This is an increase of almost 5 percentage points over the entire period.
Sources: MDRC calculations from Cleveland administrative records.
Notes: The sample includes only new recipients during their month
of sample intakes. See Table 2 for sample intake
period for each site.
While the above descriptive analyses cannot determine whether PRWORA caused some or all of these trends, they do allow the comparison of the incidence of cycling pre- and post-PRWORA. Table 13 continues this analysis by averaging the incidence of cycling across all pre- and post-PRWORA months and then calculating the difference between these averages. The table also shows the pre-and post-PRWORA averages for becoming a short-term recipient and a long-term recipient. As previously, separate estimates are presented for Cleveland and Philadelphia.
For each comparison, the post-PRWORA sample includes all new recipients who first received welfare from August 1996 through the final month of sample intake, December 1996 in Cleveland and December 1997 in Philadelphia. Results for these groups are compared to two pre-PRWORA cohorts. The first (and larger) pre-PRWORA group includes all new recipients who first received welfare from January 1993 through July 1996. (See Table 13, top panel) Since the 1996 welfare reforms were widely publicized, it is reasonable to suspect that recipient behavior — especially that of recipients who started welfare receipt closest to 1996 — may have changed prior to the actual implementation of the reforms.
To account for this reasonable change in behavior, the lower panel of Table 13 limits the pre-PRWORA sample to adults who first received welfare during January through December 1993. In other words, the behavior of recipients who began welfare receipt in 1993 is considered to be completely governed by the rules of the then current welfare regulations and not influenced by the 1996 reforms.
For the cycler outcomes in the top panel of Table 13, the PRWORA difference in becoming another type of recipient is small, but statistically significant in both sites. In Cleveland, during the pre-reform period, 11.4 percent of recipients who began welfare receipt went on to become cyclers within the next four years. In contrast, 14.8 percent of recipients who began welfare receipt during the post-reform period went on to become cyclers. This resulted (with rounding) in a 3.5 percentage point increase in the incidence of cycling between the pre- and post-PRWORA period.
The PRWORA difference in becoming a long-term recipient versus either a cycler or a short-term recipient is also statistically significant, implying more extensive changes occurred in welfare receipt behavior over the two periods. For example, the rate at which sample members became long- term recipients in Cleveland decreased by 12.7 percentage points over the pre- and post-PRWORA period. In contrast, the incidence of becoming a short-term recipient increased by 9.2 percentage points.
| Sample Intake Months | |||
|---|---|---|---|
| Site (%) | Pre-PRWORA | Post-PRWORA | Difference |
| A. Pre-PRWORA: January 1993 to July 1996 | |||
| Cleveland | |||
| Cyclers | 11.4 | 14.8 | 3.5 ** |
| Short-term recipients | 54.6 | 63.9 | 9.2 ** |
| Long-term recipients | 34.0 | 21.3 | -12.7 ** |
| Philadelphia | |||
| Cyclers | 3.8 | 6.3 | 2.5 ** |
| Short-term recipients | 47.3 | 56.6 | 9.3 ** |
| Long-term recipients | 48.9 | 37.0 | -11.8 ** |
| B. Pre-PRWORA: January 1993 to December 1993 | |||
| Cleveland | |||
| Cyclers | 10.5 | 14.8 | 4.3 ** |
| Short-term recipients | 50.1 | 63.9 | 13.8 ** |
| Long-term recipients | 39.4 | 21.3 | -18.2 ** |
| Philadelphia | |||
| Cyclers | 3.8 | 6.3 | 2.6 ** |
| Short-term recipients | 41.6 | 56.6 | 15.1 ** |
| Long-term recipients | 54.7 | 37.0 | -17.7 ** |
| Sources: MDRC calculations using state and
county administrative records. Notes: The sample includes only new recipients during their month of sample intake. See Table 2 for sample intake period for each site. There were 26,365 and 49,067 sample members included in the regressions for Cleveland and Philadelphia, respectively. Sample intake for the post-PRWORA group occured on or after August 1996. "**" indicates statistical significance at the 0.05 level or smaller. |
|||
Table 13 also presents the results for Philadelphia. The top panel of this table shows that 6.3 percent of post-PRWORA groups of new recipients became cyclers, compared with 3.8 percent of pre-PRWORA groups. That is, the incidence of becoming a cycler versus another type of recipient increased over the period. This difference of 2.5 percentage points is small, but statistically significant. This is similar to the Cleveland results. The table also shows similar differences in the incidence of short-term and long-term recipiency compared to Cleveland's results. That is, short-term recipiency in Philadelphia increased by 9.3 percentage points, while long-term recipiency declined by 11.8 percentage points. It is interesting to note that while the pre- and post-PRWORA differences in Cleveland and Philadelphia are very similar, the overall levels of cyclers, both pre- and post-PRWORA, are significantly lower in Philadelphia than in Cleveland. Furthermore, the levels of long-term recipients are significantly higher in Philadelphia than in Cleveland.
The lower panel of Table 13 shows the results of a similar analysis using the more conservative pre-PRWORA period of 1993. The panel shows the same patterns as reported in the top panel, although the decrease in the rates of long-term recipiency and increase in the rates of short-term recipiency are noticeably larger in both sites.
How large are these effects? Comparing the PRWORA differences in cycling from the Urban Change sites to the effect of the programs from the random assignment evaluation sites helps to answer this question.(41) Compared with most welfare-to-work programs studied using random assignment in Table 12 (limited to new recipients), the differences reported in Table 13 are large. For example, Table 12 revealed that Connecticut's Jobs First program reduced welfare cycling by 1.5 percentage points, increased short-term recipiency by 2.4 percentage points, and decreased long-term recipiency by 0.9 percentage points in the four years after people entered the study. None of these effects were statistically significant at the five percent level. The lack of impacts compare with Cleveland's significant differences of a 3.5 percentage point increase in cycling, a 9.2 percentage point increase in short-term recipiency, and a decrease in long-term recipiency by 12.7 percentage points. While the differences in Table 13 are not causal, they do suggest that PRWORA may have influenced the large changes in recipient behavior.
[ Go to Contents ]
(38) Recall that new recipients are defined as sample members who were just starting a welfare spell around their time of sample intake.
(39) A strong economy increases the demand for workers in the short term and a sustained, strong economy increases the demand for lower-skilled workers, resulting in higher employment at all skill levels. Declining caseloads suggest people are leaving the welfare rolls faster than new applicants start.
(40) Because the data for Cleveland extend back only to July 1992, there is no way to know whether someone received benefits prior to July 1992. The groups of "new" recipients, as they are defined here, may contain many people who had received benefits prior to July 1992 but who had not received benefits between July 1992 and the month when they began receiving benefits anew. Later groups of "new" recipients are likely to contain fewer relatively recent welfare recipients and more truly new recipients.
(41) Recall that this analysis, presented in Table 12, seeks to answer whether the program implemented at each evaluation site affected the likelihood of cycling.
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)