A Profile of Families Cycling on and off Welfare. What are the employment and welfare outcomes after cycling?


As noted in the Introduction, cyclers may eventually attain stable employment without welfare, similar to welfare leavers who did not cycle. To explore this view further, we compare employment and welfare outcomes of cyclers, short-term recipients, and long-term recipients in the five-year follow-up sample during the year following the observation period.

Table 11 presents adjusted means of employment and welfare outcomes for the fifth year after sample intake. These calculations were performed for the five-year follow-up sample.(36) As with the findings in Table 8, the comparisons of employment and welfare outcomes control for differences among the three groups in background characteristics (displayed in Table 6,) that could affect sample members' patterns of work and welfare after sample intake.

Table 11.
Differences Among Cyclers, Short-Term Recipients, and Long-Term Recipients
In Selected Employment and Welfare Outcomes
During Year 5 After Sample Intake
Outcome(%) Cyclers Short-Term Recipients Difference Cyclers Long-term Recipients Difference
Ever employed in year 5 65.9 59.5 6.4 ** 65.9 60.6 5.3 **
Employed during all four quarters of year 5 28.1 31.1 -3.0 ** 28.1 25.8 2.2 **
  If employed in year 5 42.9 52.3 -9.3 ** 42.9 42.2 0.7
Employed in the last quarter of year 5 41.3 39.5 1.9 ** 41.3 39.0 2.3 **
  If employed in year 5 63.2 66.7 -3.6 ** 63.2 63.7 -0.6
Received welfare in the last month of year 5 25.0 7.4 17.5 ** 25.0 33.5 -8.6 **
Employment and welfare status in the last quarter of year 5
  Employed and did not receive welfare 29.6 36.6 -7.0 ** 29.6 25.4 4.2 **
  Employed and received welfare 11.8 2.9 8.9 ** 11.8 13.6 -1.9 **
  Not employed and received welfare 18.9 5.0 13.9 ** 18.9 23.7 -4.7 **
  Not employed and did not receive welfare 39.7 55.5 -15.8 ** 39.7 37.3 2.4 **
Sample size 9,195 53,237   9,195 83,719  
Sources: MDRC calculations from state and county administrative records.
Notes: Calculations were performed for sample members with five years of follow-up data. The samples were equally weighted by site. The Connecticut Jobs First sample lacked five years of follow-up data and was excluded from the calculations.
Estimates are regression-adjusted using ordinary least squares, controlling for sample member characteristics and environmental conditions. Results from logistic regressions performed on binary outcomes were consistent with the above results.
"**" indicates statistical significance at the 0.05 level or smaller.

The results in Table 11 point to several positive longer-term trends. First, employment levels for each group converged: Between 60 and 66 percent of members of the five-year follow-up sample worked for pay during year 5 and about 40 percent of each group were still employed during the last quarter of the year. In addition, the large majority of sample members  even long-term recipients  were no longer receiving welfare benefits at the end of year 5. Finally, each group demonstrated greater self-sufficiency during year 5 compared with the observation period. More specifically, it may be recalled that a larger percentage of cyclers who found employment during the observation period received earnings and welfare benefits during the same quarter than relied on earnings alone. This pattern was reversed by the end of year 5, when 29.6 percent of cyclers were employed without welfare, in contrast to only 11.8 percent who received both earnings and welfare benefits. A similar reversal occurred among long-term recipients.

Less positively, the table also shows that joblessness and unstable employment remained a problem for all three groups. About 35 to 41 percent of sample members never worked for pay during year 5. Furthermore, one-third or more of sample members ever employed during year 5 were no longer working during the last quarter.

Cyclers continued to show a higher incidence of employment than either short-term or long-term recipients in year 5. However, employment among cyclers was not as stable as that among short-term recipients. Specifically, 28.1 percent of cyclers were employed during all four quarters of year 5, compared with 31.1 percent of short-term recipients. This result continues to hold among those employed in year 5. Among employed sample members, the proportion of cyclers who worked for pay during all four quarters of year 5 (42.9 percent) was more than 9 percentage points below the level for short-term recipients. While cyclers had more stable employment compared with long-term recipients (cyclers are employed 2 percentage points more than long-term recipients during all four quarters), among those employed, there was no difference in employment stability in year 5.

Again, cyclers exhibited stronger attachment to the welfare system than short-term recipients, but less attachment than long-term recipients, as represented by the percentage that received TANF benefits in the last month of year 5. In this month, cyclers were 17.5 percentage points more likely than short-term recipients to be receiving cash assistance, but 8.6 percentage points less likely than long-term recipients.

As noted above, nearly 30 percent of cyclers worked for pay at the end of year 5 and did not receive welfare benefits. This average exceeded the level for long-term recipients by 4 percentage points but was 7 percentage points below the level for short-term recipients. Cyclers were also about four times more likely than short-term recipients to combine work and welfare (11.8 percent to 2.9 percent)  and did so only slightly less often than long-term recipients. In other words, fewer cyclers than short-term recipients were able to sustain themselves with employment alone.(37)

Thus, by the end of year 5, cyclers did not "catch up" to short-term recipients in key measures of employment stability and self-sufficiency, although they continued to fare better than long-term recipients.

In summary, cyclers were shown to be a group in the middle  less disadvantaged in the labor market than long-term recipients, but less able than short-term recipients to attain stable employment and work without welfare. Furthermore, cyclers were the most likely to be parents of toddlers and preschoolers and therefore had the greatest need for reliable child care while employed. However, compared with short-term recipients, cyclers had less access to financial and other support from a spouse or partner.


(24) Appendix Table 1 presents the characteristics of sample members by site.

(25) In other words, the significance test measures that there is a difference between at least two of the groups, but does not indicate whether cyclers are significantly different from both short-term and long-term recipients. See NOTES in Table 5 for results of tests of differences between cyclers and short-term recipients and between cyclers and long-term recipients. A statistical test was not performed on differences in the percentage of families whose youngest child was aged less than six years.

(26) See Appendix Table 2A for the corresponding analysis by site.

(27) This finding pertains only to members of the three evaluation sites. Data on educational attainment were unavailable for the Cleveland and Philadelphia samples.

(28) See Appendix Table 2B for the corresponding analysis by site.

(29) Possibly, employment levels for short-term recipients were most subject to underreporting because of movement out of state.

(30) The average quarterly employment rate shows the percentage of quarters employed. The rate is calculated as total quarters employed divided by 16 (the number of quarters in the observation period), expressed as a percentage.

(31) See also Ver Ploeg, 2002, Table 13-15 and pp. 445-446; and Miller, 2002, Figure 1C and pp. 24, 27. Ver Ploeg found that long-term recipients averaged more than both short-term recipients and cyclers in income from personal earnings, welfare, and food stamps, measured with administrative data. Miller, using survey data on total household income (which includes earnings of other household members) found that welfare leavers in Connecticut Jobs First, Florida FTP, and Vermont WRP averaged higher incomes than longer-term stayers.

(32) See, for example, Hamilton et al., 2001, Appendix Tables F.1-F.3, pp. 392-395.

(33) Earnings from other family members are not included, nor are Earned Income Tax Credits (EITC) that can supplement personal earnings.

(34) See Table 2 for information on intake dates and sample sizes for survey respondents. Responses were pooled and weighted equally by site. Within each site, the number of cyclers is relatively small. Therefore, simple means and percentages are shown in Table 9. These measures were not adjusted for differences in respondents' characteristics, nor for site differences in labor market and other environmental conditions.

(35) These differences are not solely the result of higher employment levels for short-term recipients at the time of interview. When only respondents who were working at interview are considered, about 56 percent of short-term recipients, 43 percent of cyclers, and 39 percent of long-term recipients reported working for an employer who offered medical coverage.

(36) Note that this subsample does not include the Connecticut Jobs First program. See Table 2 for the composition of the sample.

(37) It is also worth noting that short-term recipients had, by far, the highest percentage of sample members who did not work for pay and did not receive welfare. Prior research has shown that some people in this status are actually working out of state or in jobs not covered by the Unemployment Insurance system, whereas others are truly unemployed. Most likely, the group without employment or welfare benefits included sample members who were living with a spouse, partner, or other adult who was employed and providing financial support to the sample member and her children. However, others may have had no steady source of income and were living in extreme hardship.

View full report


"report.pdf" (pdf, 369.93Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®