Hall of Mirrors — expanded estimators
Archive index · Regression tests · Results & settings (JSON) · Methods & code
Bertrand et al. (2021), QJE. https://doi.org/10.1093/qje/qjab023
1 specification. Effects are in outcome standard deviations; intervals are 95%.
Treatment adoption
Hall of Mirrors: 100 sampled trajectory patterns, not individual pairs. Exact frequency weights represent the underlying pair panel.
HALL-MIRRORS-F1-COCOMMENT
38,528 observations · 3,253 units · 3,253 clusters · 226 treated / 3,027 control units.
Estimate comparison
| Estimator | Estimate (SDs) | SE | 95% CI | Averaging |
|---|---|---|---|---|
| Static TWFE | 0.247 | 0.024 | [0.200, 0.294] | Coefficient |
| Dynamic TWFE | 0.147 | 0.032 | [0.084, 0.211] | equal_periods |
| BJS / did2s | 0.260 | 0.024 | [0.214, 0.307] | equal_periods |
| Sun–Abraham | 0.161 | 0.032 | [0.099, 0.222] | equal_periods |
| CS-DID (equal event-time) | 0.123 | 0.032 | [0.061, 0.185] | Equal post-event times |
| CS-DID (treated cells) | 0.160 | 0.031 | [0.098, 0.221] | Treated cells |
Event studies
Periods −10 to +10; pointwise 95% intervals where available.
