didq: A command for treatment-effect estimation under alternative assumptions

Abstract

When several pretreatment periods are available, identification of the treatment effect in a difference-in-differences framework requires an assumption relating dynamics for controls and treated in absence of treatment. Mora and Reggio (2012, Working Paper 12-33, Universidad Carlos III de Madrid) define a family of alternative identifying assumptions and propose a model that, contrary to the usual econometric specifications, allows one to identify the treatment effect for any given assumption in the family. In this article, we introduce a command, didq, that implements the model presented in Mora and Reggio, reports the estimated effect under alternative assumptions, and performs tests for the equivalence of the estimates. We also explain how to use the command to obtain the standard difference-in-differences estimator with or without polynomial trends.