to the enactment of PRWORA to outcomes of a cohort of leavers who left welfare after enactment of PRWORA. Both of these designs have weaknesses in drawing causal conclusions.1
Factors outside of welfare, such as the economy, may also change and affect the outcomes of welfare leavers, making it difficult to assess whether outcome changes are due to the reforms or to the other factors using these methods. Another weakness of these methods is that the characteristics of the people leaving welfare at the time of the study, or at the time the cohorts are drawn, may be driving changes in outcomes. For example, if a cohort of leavers is drawn when the caseload is relatively small, the leavers may be comprised primarily of those who have the most barriers to leaving welfare, such as substance abuse, very young children, or little work experience. Their outcomes after leaving may be much different than the outcomes of a cohort of leavers drawn when the caseloads are relatively large, since this cohort may be composed of leavers with fewer barriers to self-sufficiency. This second problem of the composition of the caseload is also a problem even if the leaver studies are only used for monitoring, and not evaluation, purposes. For example, a monitoring study may be conducted to roughly quantify the need for child care services of those who leave welfare. Those who leave welfare in a time when caseloads are just beginning to drop may be able to leave because they had an easy time securing child care, while those who could not easily find childcare may not leave welfare until much later. It would be hazardous to base conclusions about the need for childcare from any single cohort of leavers if one does not know much about that cohort of leavers.
The National Research Council report (1999) suggested that as a crude means of standardizing descriptions of the caseload and the outcomes of leavers across time and across areas, outcomes could be stratified by the past welfare receipt history and past work experience of welfare leavers. Standardizing the composition of the caseload and the groups of the leavers would then make comparisons of outcomes of leavers across time and jurisdictions more credible because leavers with similar work and welfare receipt histories would be compared to each other. The purpose of this paper is to classify characteristics of welfare leavers and stayers and their outcomes by their preexit benefit receipt and employment experiences to illustrate one method the leaver studies might use to standardize their results to make comparisons across time and jurisdictions more credible. No attempts to make causal attributions are made in this study.
The second section of this study describes the data used. The third section examines the past welfare receipt, employment, and earnings histories of the caseload of AFDC recipients in 1995. Section 4 examines whether and how much welfare leavers work and earn after leaving, whether they return to welfare or use