never have been on welfare and so would not be able to answer questions about welfare receipt or even some of the child's case history.
Those who planned to include the child-only cases asserted that they could be considered another subgroup for analysis and that the additional cost of keeping these cases in the administrative data sets is minimal. Special survey designs could be developed for child-only cases. One workshop participant suggested that a family and its budget is really the focus of these studies and, therefore, in the sense that child-only cases are benefit reduction cases, the behavioral response to this change is of great interest. Another reason for studying this population is to understand the behavioral response of child-only cases whose parents are immigrants. As immigration rules are being more strictly enforced, many immigrant child-only cases may be closed because the parents fear being reported to the Immigration and Naturalization Service. Another reason to include child-only cases is that child-only cases are a growing segment of the caseload and, especially as time limits take effect, it will be important to know the composition and behavior of this group.
Many workshop participants stressed the need for additional studies of the reasons for the increase in child-only cases. The work requirements for these groups vary considerably by state. The number of child-only cases in the states due to immigrant status also varies across states. Understanding these variations would be beneficial to understanding the well-being of these families and to developing policies about them.
Workshop participants also emphasized the importance of collecting data on the program participation histories of leavers. There are several types of program participants: many may be first-time leavers; others may have left welfare before and later returned to the rolls. In order to characterize the composition of the leaver population, it would be useful to know who has left, how long they had received benefits, if they had previously cycled on and off of welfare, what benefits and the levels of benefits that they previously received, and their labor force histories. A careful analysis of outcomes stratified by these characteristics would help make comparisons across states and jurisdictions and would help control for differences in caseloads among the jurisdictions.
EVALUATION METHODOLOGIES
Discussion at the workshop on what evaluation methodologies would be appropriate for the 14 studies of welfare leavers focused on the description and monitoring purposes of the grants, in contrast to the evaluation purpose of the grants. Most participants agreed that these studies could adequately describe and monitor what is happening to the welfare caseload, but that evaluating the effects of the new policies with these studies will be more problematic.
There are several reasons why the cohorts defined in the 14 studies may not be comparable for policy evaluation. First, it is important to remember that economic conditions change over time. The local economic conditions faced by the leavers at one time may be very different than the economic conditions faced by leavers at another time. Second, the types of people who leave the rolls at two different times may be different. For example, cash assistance recipients who have greater labor market skills may leave earlier than those who have few labor market skills. If this is true, comparisons of an early cohort of leavers to a later cohort of leavers should be done with recognition of the differences in the caseload composition. Workshop participants emphasized that monitoring what has happened to these two cohorts will be useful for identifying potential ways that different groups may be underserved or what services are needed by those who leave welfare. It was