A1408
Title: Potential outcome modeling and estimation in DiD designs with staggered treatments
Authors: Siddhartha Chib - Washington University in Saint Louis (United States) [presenting]
Kenichi Shimizu - University of Alberta (Canada)
Abstract: The aim is to propose the first potential outcome modeling of difference-in-differences designs with multiple time periods and variation in treatment timing. Importantly, the modeling respects the two key identifying assumptions: parallel trends and no-anticipation. A straightforward Bayesian approach is then introduced for estimation and inference of the time-varying group-specific Average Treatment Effects on the Treated (ATT). To improve parsimony and guide prior elicitation, the model is reparametrized in a way that reduces the effective number of parameters. Prior information about the ATT is incorporated through black-box training sample priors and, in small-sample settings, by thick-tailed t-priors that shrink ATT of small magnitudes toward zero. A computationally efficient Bayesian estimation procedure is provided, and a Bernstein-von Mises-type result is established that justifies posterior inference for the treatment effects. Simulation studies confirm that the method performs well in both large and small samples, offering credible uncertainty quantification even in settings that challenge standard estimators. The practical value of the method is illustrated through an empirical application that examines the effect of minimum wage increases on teen employment in the United States.
