Although federal welfare reform was implemented July 1, 1997, many believe that there was an anticipatory effect, and consequently, some AFDC recipients in the early months of 1997 voluntarily stopped their cash assistance although still eligible. Others also believe that potentially eligible recipients avoided applying, believing, mistakenly, that welfare "as we knew it" was a thing of the past. With this in mind, we wanted to select a cohort of recipients who entered AFDC, Medicaid, or foster care near enough to welfare reform to avoid the effects of the changed policies in each state, but not be influenced by any anticipatory effects. We chose children who entered one of the three programs in calendar years 1995 and 1996 and followed them through the end of 1996.
We also wanted a study population that shared similar program experiences. Such shared experiences, such as entry into a program during the same period of time, allow us to control for factors that might affect outcomes. A population of children with entry dates from different historical periods, such as that of a point-in-time population, could also fundamentally change the potential distribution of the likelihood of the events of interest occurring. Understanding factors that affect the flow of children among programs in either case would be difficult. Choosing a point-in-time population would have included children with widely varying durations of participation, as well as varying set of previous programs used. A child in care on a particular day may have been in a particular program for any duration between one day or 17 years. Therefore, the reasons for children entering foster care when they had been receiving AFDC for a considerable period before 1995 would be confounded. We would not know if children were entering foster care because of long-term poverty or because of changes in parenting during a short period of time before foster care entry. Also, we find that children are more likely to enter foster care at the beginning of their AFDC career and that the likelihood decreases over time.
In addition, for the point-in-time population of some programs, we would never expect the majority of the children to leave a program. Another study population might be an exit cohort. Choosing an exit cohort excludes those children who had not formally left the programs, but were nevertheless at risk of entering one of the other programs.(3) Although a transition from foster care to AFDC requires a formal opening of an AFDC case, it does not require a closing of the Medicaid case. Therefore, if we only look at children who exited AFDC, we would miss a transition to foster care of a child receiving AFDC. Because of these issues, one should be careful when comparing the findings in this study to studies with either repeated cross-sections or the "leaver" studies.
In an "entry" cohort study such as the one presented in this report, it is imperative to understand any possible effects of the study population's experiences prior to the "observation" period. In studies that use similar longitudinal methods similar to ours, it is well documented that experiences of the new entry cohorts (those who enter a particular program for the first time) are often quite different from those who had previous service experiences (i.e., re-entries or those who had previous contacts in other service systems). To control for the effect of the previous service experiences, to the extent the longitudinal service data allow, we include those children with no AFDC, Medicaid or foster care two years prior to AFDC entry.
We believe that this study population is the most appropriate one for accurately portraying the movement of children from income maintenance receiving families into foster care. There is a high representation of young children our AFDC group and that is also true of the children that enter foster care. Foster care in the larger states in the U.S. has become a program that targets children under the age of five (Wulczyn et al., 1999).