Using Weights in Two-Way Fixed Effects and Event Study Regressions

Recent papers show that coefficients in two-way fixed effect (TWFE) and event study (ES) regressions can hardly be interpreted as the average treatment effect, depending on heterogeneity in the treatment effects. This paper proposes to contrast estimates with and without weighting to detect heterogeneity in treatment effects. If estimates with and without weighting are similar to each other, TWFE and ES regressions are enough to identify average treatment effect. However, if they are different, it is necessary to use alternative estimators robust to heterogeneous effects to identify the average treatment effect (e.g., de Chaisemartin and D'Haultfuille, 2020). For an empirical application, I revisit literature estimating the impact of unilateral divorce law on the divorce rate with US panel data. The discrepancy between estimates with and without weighting implies that the divorce law affected divorce rates differently across groups and time. After using the alternative estimators, the effect of this law on the divorce rate is smaller than previous papers found