This paper describes the design of a study intended to examine the devolution of welfare in four cities. The study will measure the economic effects of new policies on individual and potential welfare recipients; describe the new programs and policies as they are put in place; examine the experiences of a small number of families; look at how the new policies and funding mechanisms affect both for-profit institutions and nonprofit and public service delivery systems; and assess changes in the social and economic vitality of neighborhoods. The study focuses on the portion of the impact analysis that uses administrative data. It uses a multiple cohort design, in which a group of those receiving welfare or at risk of receiving welfare will be followed over time, and their outcomes will be compared. The paper describes: a model of the impacts of welfare reform (the treatments and predicted effects); the proposed design (a multiple cohort analysis); hypothetical examples of multiple cohort analysis; random assignment evaluation as multiple cohort analysis; national variation in outcomes across cohorts; selecting cohorts; and data sources. An appendix presents econometrics of the multiple cohort design. (SM)