An Evaluation of Difference-in-Differences Methods Using Placebo Event Studies

Researchers are faced with the choice of which of the many recently developed difference-in-differences methods to use in practice. To assess these estimators' relative performance for single-unit event studies, we conduct 134,000+ state-level placebo event studies across 13 estimators. We find that no single method dominates. Performance is context-dependent, with synthetic-control-like methods sometimes outperforming and sometimes underperforming two-way-fixed-effect-like and matching methods. Performance also varies across states at least as much as it does across estimators. Our results highlight the need for practitioners to conduct placebo tests to understand the performance of methods in their research context.