CHARACTERIZING SELECTION BIAS USING EXPERIMENTAL-DATA

Semiparametric methods are developed to estimate the bias that arises from using nonexperimental comparison groups to evaluate social progra ms and to test the identifying assumptions that justify matching, sele ction models, and the method of difference-in-differences. Using data from an experiment on a prototypical social program and data from none xperimental comparison groups, we reject the assumptions justifying ma tching and our extensions of it. The evidence supports the selection b ias model and the assumptions that justify a semiparametric version of the method of difference-in-differences. We extend our analysis to co nsider applications of the methods to ordinary observational data.