The information presented here is drawn from an archive containing full histories of all AFDC grant cases from 1990 through 1995. The data points analyzed here are based upon an extraction of records for all children who were active recipients in an AFDC grant case in any of six successive annual June cross-sections. The history of the grant unit itself can be traced fully between these June time-points, but the actual attachment of the individual to the grant during the unobserved time must be inferred. The use of annual cross-sectional pulls inevitably implies that some child participation in the AFDC program is missed entirely, such as when a grant is both opened and closed during the 10-month time period between any July and the following May. Given certain constraints, we believe that the group of cases developed for this analysis presents a relatively accurate picture of the Illinois child AFDC population. As will be seen in the following analysis, even though we are missing some of the rapid "on-and-off" movement that occurs with some welfare participants, there is a marked overall continuity and stability in the welfare histories of the children we observe.
For each AFDC-recipient child, we can identify their age, ethnicity, geographic area, gender, program participation (Medicaid, Food Stamps), and relationship to the official grantee. We can also identify the same information for the grantee, all recipients in the grant unit, and some other persons associated with the grant unit. Because individuals often move between grant units, and grant units often split apart or reform into new grant units, a careful unduplication of child records was pursued to insure that each child history is complete (to the extent possible), and that the same child does not appear multiple times in the data just because he/she received benefits under different grant units. For analysis, children were restricted to persons between the ages of 0 and 17. Even if grant support continued beyond the eighteenth birthday, 18-year-olds were excluded from this analysis, because the custodial nature of care is redefined when a child reaches the age of majority.
A "relation to the grantee" field was used to obtain a preliminary classification of children into "parent" cases, "relative" cases, and "other" cases. All relationships coded as parent, mother, father, step-parent, etc. were classed as parent cases, and those classified as grandparent, aunt/uncle, sibling, cousin, etc. as relative-only cases. After this initial classification, each relative case was screened specifically for the presence of a "credible mother" in the grant unit.(17) For the purposes of analyzing kinship care, we assumed that classification errors in the direction forcing a few "relative-only" cases into the "parent" population should have much smaller potential corrupting effect on our conclusions than would result from classifying a number of parent-present cases into the "relative-only" population. Not only are "relatives" are the focus of study, the pool of relative cases is also much smaller than the pool of parent cases.(18)