In policy evaluations, the standard difference-in-differences (DID) method assumes a constant average confounding effect of concurrent events for the comparison group unaffected by the policy and the experimental group affected by the policy. We estimate a pair of prognostic scores representing the predicted pre-policy outcome and the predicted post-policy outcome under the comparison condition. Subsequent DID analyses within subclasses defined by the prognostic scores allow for a calibrated adjustment. We evaluates the performance of alternative DID strategies through simulations and illustrate with an evaluation of a system-wide policy.
Hong, G., Nomi, T., & Yu, B. (2012). Prognostic score-based difference-in-differences. In JSM Proceedings, Social Statistics Section. Alexandria, VA: American Statistical Association, 4952-4966.