Abstract

This paper proposes an estimator for the average treatment effect on the treated (ATT) in difference-in-differences (DID) research design with the method of covariate balancing propensity score (CBPS). The proposed estimator can achieve the semiparametric efficiency bound when both propensity score and outcome regression working models are correctly specified and consists with the true ATT even when one of the two working models is misspecified, as with the alternative DID estimators. However, the proposed DID estimator is robust for inference under one working model misspecification scenario and outperforms the alternative DID estimators when both models are locally misspecified. We evaluate the finite sample performance of the proposed estimator via simulation.