We conduct a meta-analysis of causal panel analyses under parallel trends using two-way fixed effects (TWFE) models published in three top journals in Political Science (APSR, AJPS, JOP) from 2017 to 2023 (Paper). Using data from 49 studies, we find:
- Modern DID estimators, which are designed to be robust to heterogeneous treatment effects (HTE), yield qualitatively similar but highly variable results to TWFE estimates.
- While a few studies show clear signs of parallel trends (PT) violations, many lack evidence to support this assumption.
- Many studies are underpowered when accounting for HTE and potential PT violations.
Here, we present replication results for each paper we reanalyzed.