A Propensity Score Model of Costs and Benefits Related to Army Service

A major obstacle to the establishment and measurement of valid cause and effect relationships between Army service and the costs and benefits of such service is the impossibility of performing experiments with random assignment of individuals to service and non-service groups. A recent innovation in statistical theory, propensity score methodology, allows one to draw from an observational (or retrospective) study virtually the same kind of causal inferences that previously have been possible only with experimental studies. Propensity score methodology has been successfully used in medical, social science, and educational contexts. Phase I will develop a specimen propensity score model of costs and benefits of Army service, will identify and specify the data sources necessary to implement and test the model, and will determine the procedures and scope of the effort needed to test the model. Phase I will include a preliminary specification of the variables appropriate for use in the final model. Attention will be given to the question of oversampling women and minorities so that separate models may be developed for them. Not only are such separate models desirable heuristically, they are also particularly appropriate when using propensity scores. Phase II will be devoted to the elaboration, detailed specification, testing, and evaluation of the model.