Download
  • Draft Coming Soon

Abstract

In this paper, we propose a new approach to identify the distributional treatment effect parameters in differences-in-differences based on a combination of a random coefficients model with deconvolution. These treatment effect parameters depend on the joint distribution of potential outcomes, which is almost never point identified. We show that it is possible to circumvent identification of the joint distribution of potential outcomes under some conditions. Our main assumption for nonparametric identification is that the residual trends provide no information about treatment effects. We study the proposed framework in an application to an employment training program. Finally, we propose a non-parametric estimator and investigate finite sample properties through Monte Carlo simulations.


Citation

Corcuera, Paul, and Song, Kevin. 2023. “Estimating the distribution of treatment effects in difference-in-differences.” Working Paper.