Case study: highly imbalanced covariates barely move DiD estimates, and why

analisereal · x · 2026-09-17

Using the Callaway & Sant'Anna (2021) example, the author demonstrates use case (1): the estimate is -2.77% without covariates and -3.66% with them. The observed covariates are highly imbalanced, yet including them barely changes the estimates — the decomposition explains why, showing how observed confounders (used under a conditional PT assumption) affect results.

Related event: New framework quantifies confounding strength in DiD sensitivity analysis(4 posts)→

Original post →

More from Research

Research channel →