This report examines participation rates in the Aid to Families with Dependent Children (AFDC) program. TRIM2, a microsimulation model that simulated the eligibility and benefit rules of the AFDC program on a state-by-state basis, showed that there had been a dramatic decline after 1981 in the rate at which AFDC families were applying for and receiving benefits. This phenomenon was caused by the change in the coding of subfamilies (adults who live with children but are not their biological parents) by the Census Bureau on the Current Population Survey, which is the basic input file used to compute eligibility. There was no massive behavioral change on the part of AFDC eligibles in the early 1980s: between 150,000 and 400,000 subfamilies (depending on the year) were omitted from the eligibility estimates between 1967 and 1980. Comparative analyses of the AFDC program that encompass the period before and after the coding change might be impaired by the impact of the change. Participation rates are much lower than previously believed, and are perhaps not at a saturation level. Analysis of trends in the characteristics of AFDC eligibles from 1979 to 1984 reveal the following changes after 1980: (1) more subfamily and family heads were likely to be black, female, young and never married; (2) more family and subfamily heads lived in higher income households and contributed to a lower percentage of household income; and (3) more subfamily heads had a child under 6 years of age. Implications for policy and research are discussed. Data are presented on 14 tables and figures. A bibliography is included. (BJV)