Leavers, Stayers, and Cyclers An Analysis of the Welfare Caseload. Samples


For the analysis in this report, the samples in each program are restricted to people for whom data were available for at least 36 months but no more than 60 months after they were randomly assigned into a program. The evaluations have varying lengths of follow-up, and this restriction is imposed so that “staying on welfare” and “leaving welfare” mean something similar across programs. Some people in NEWWS, for example, are followed for up to six years after random assignment, while people in MFIP are tracked only for four years. 
Table 1 presents the sample sizes, by project, for the full administrative records sample and the smaller survey sample. The full sample is used for analyses of demographic characteris-tics and potential barriers to work, and the survey sample is used for analyses of surveyoutcomes, including household income, household composition, and material hardship.Each analysis was run for the pooled sample and then for each program separately, and the pooled analyses were weighted to reflect the size of the state’s welfare caseload in the mid-1990s. Although the data are weighted, the sample should not be considered a representative sample of the recipient population in these states, owing to the idiosyncrasies of the evaluations. Some programs, for example, oversampled new applicants relative to ongoing recipients, while others screened out exempt individuals prior to random assignment. In addition, each of the evaluations operated in selected counties within the state involved.
When the programs are analyzed individually, they are presented in an order that roughly corresponds with the strength of their mandate to leave welfare. The first group is incentives only and includes the two programs that used enhanced financial incentives relative to the traditional AFDC program but did not impose any work requirements (WRP Incentives Only and MFIP Incentives Only). The second group includes programs that combined financial incentives with a work requirement (the full WRP program and the full MFIP program). The third group includes traditional welfare-to-work programs that did not use financial incentives but that encouraged recipients to leave welfare either by offering them education-focused services (NEWWS, education-focused) or employment-focused services (NEWWS, employment-focused). The final group includes the two time-limit programs, both of which also offered enhanced financial incentives (Jobs First and FTP). 
Most analyses in the report are restricted to people in the program (or experimental) groups in each evaluation. These are the most relevant groups, since members were subject to the reforms being tested, many of which are key elements of states’ new TANF programs. In addition, results are presented for a pooled sample - combining data from each evaluation - and for each evaluation separately. The control groups are used only toward the end of the report, when the program and control groups are compared in order to infer the effects of the particular program being tested. 
Table 1
Sample Sizes, by Program
Project Administrative Records Sample Survey Sample
WRP, Incentives Only 2,212 587
MFIP, Incentives Only 1,621 567
WRP 6,591 613
MFIP 3,936 1,280
NEWWS, Education focused 12,839 2,830
NEWWS, Employment focused 7,659 1,763
Jobs Firs
674 89
FTP 917 748
Total 36,449 8,477
SOURCE:  Data from MDRC evaluations of the following programs:
     WRP (Vermont's Welfare Restructuring Project) 
     MFIP (Minnesota's Family Investment Program) 
     NEWWS (National Evaluation of Welfare-to-Work Strategies) 
     Jobs First (Connecticut's Welfare Reform Initiative) 
     FTP (Family Transition Program) 

2The results using the full sample to compare demographic characteristics and barriers to work were similar when restricted to the smaller survey sample. 

View full report


"Miller-MDRC-02.pdf" (pdf, 205.87Kb)

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®