Study Significance
A large replication and reanalysis of 49 articles from leading journals that used two-way fixed effects (TWFE) models on observational panel data with binary treatments. The project responds to recent methodological concerns about heterogeneous treatment effects (HTE) and violations of the parallel trends (PT) assumption, alongside a proliferation of new estimators and procedures that have confused best practices.
Scope and Data Reanalyzed
They replicated and reanalyzed 49 published studies that used TWFE models on observational panel data with binary treatments; the project reused the original studies' data and designs where available to apply alternative estimators and diagnostics.
Comparison Across Estimators and Diagnostics
For each study they applied six HTE-robust estimators, ran diagnostic tests aimed at detecting parallel-trends violations, and performed sensitivity analyses to assess robustness to alternative approaches.
Main Findings and Interpretation
HTE-robust estimators often produce qualitatively similar conclusions to the original TWFE results, but point estimates and precision vary substantially across estimators; a minority of studies show clear signs of parallel-trends violations while many studies present no evidence that the PT assumption holds, and accounting for HTE and PT concerns suggests many original studies are underpowered.
Implications for Research Practice
The authors emphasize the importance of strong research designs and routine validation of key identifying assumptions: for example, justifying estimator choice when HTE is plausible, reporting diagnostic checks for parallel trends, and assessing statistical power under HTE-aware models before making strong causal claims.
