Abstract
We develop a high-dimensional covariate balancing propensity score (HCBPS) approach for estimating the average treatment effect on the treated (ATT) within a difference-in-differences (DID) research design. This method accommodates, but does not necessarily require, the dimension of covariates to exceed the number of observations. We show that the proposed HCBPS DID estimator possesses several desirable properties: (i) local efficiency, (ii) double robustness in terms of consistency, and (iii) double robustness in terms of inference. Simulation studies support the proposed estimator's desirable finite-sample performance.