Power implications of heterogeneity-robust DiD versus TWFE

Determine whether the flexibility of heterogeneity-robust Difference-in-Differences and event study estimators designed for short panel data necessarily entails a meaningful loss in statistical power compared to two-way fixed effects regression estimators, across empirically relevant designs with heterogeneous treatment effects.

Background

Modern Difference-in-Differences (DiD) and event paper estimators allow for heterogeneous treatment effects and work with short panels, but empirical applications often report wider confidence intervals relative to more restrictive two-way fixed effects (TWFE) models. The paper highlights that it is unclear if this loss of precision is inherent to the flexibility of heterogeneity-robust methods or a consequence of non-optimal weighting or estimator choice.

The authors develop semiparametric efficiency bounds and efficient influence functions to address precision concerns, but they explicitly flag the general question of whether modern DiD necessarily sacrifices power relative to TWFE as open. This question matters for empirical practice and methodological guidance when selecting estimators under heterogeneous effects.