A Regression-Based Strategy for Defining Subgroups in a Social Experiment

Two prominent themes often emerge from evaluations of education and social program evaluations: (1) the interventions being studied serve diverse populations, even if they are intended to target groups with particular characteristics; and (2) the interventions' impacts vary across groups within the population being served. Thus, most evaluations of education, employment and training, welfare-to-work, and health-related interventions are not only interested in the question "What works?" but in the question "What works for whom?" This leads to an extensive investment in subgroup analysis. This paper describes a "regression-based" strategy for defining subgroups that is more systematic than traditional "accumulation" strategies. It relies on data and methods used in the Manpower Demonstration Research Corporation's Career Academies Evaluation to illustrate the use of this technique and discusses its various advantages and limitations. (SLD) Reproductions supplied by EDRS are the best that can be made from the original document. MDRC Working Papers on Research Methodology A Regression-Based Strategy for Defining Subgroups in a Social Experiment James J. Kemple and Jason C. Snipes