DiD: Event study (dynamic effects) reported

What to Check

A static two-period DiD estimate collapses the entire treatment effect into a single number, hiding the time path of effects and masking pre-treatment trends. An event study plots the coefficients on leads and lags of treatment, serving three purposes: (1) validating parallel pre-trends, (2) detecting anticipation effects (treatment effect starting before treatment), and (3) characterizing whether effects are permanent, temporary, or growing.

This test is distinct from did-parallel-trends-plot, which checks only whether any pre-trend visualization exists. This test requires a full event study with both pre- and post-period coefficients.

How to Check

1. Look for a figure or table presenting event-study (dynamic treatment effect) estimates: coefficients β_{t-k} and β_{t+k} for k = 1, 2, … periods before and after treatment.
2. Check that pre-treatment coefficients (leads, t−1, t−2, …) are shown and are statistically indistinguishable from zero. The omitted period (normalized to zero) is typically t−1 or the last pre-treatment period.
3. Check that post-treatment coefficients (lags, t+1, t+2, …) are shown, allowing the reader to assess treatment effect dynamics.
4. For staggered treatment designs (multiple treatment cohorts): verify that the event study uses a heterogeneity-robust estimator (Callaway & Sant’Anna 2021, de Chaisemartin & D’Haultfœuille 2020, or Borusyak et al. 2024) rather than a two-way fixed effects (TWFE) event study, which is biased under treatment effect heterogeneity.
5. Event study confidence intervals should be reported (usually 95% CIs). A “joint test” of pre-trends (F-test of all pre-period coefficients = 0) is informative but optional.

Pass Condition

Event study coefficients for both pre- and post-treatment periods are plotted or tabulated. Pre-treatment coefficients are close to zero. The estimator is appropriate for the design (heterogeneity-robust if staggered).

Failure Examples

1. Static estimate only: Paper estimates a two-period DiD and reports a single ATT estimate. No dynamic effects shown. The reader cannot evaluate parallel trends or treatment dynamics. Fails.
2. Pre-periods only: Paper shows a pre-trend plot (satisfying did-parallel-trends-plot) but reports no post-period dynamics. Fails this test (passes did-parallel-trends-plot).
3. TWFE event study in staggered design: Paper has staggered rollout across 5 cohorts but runs a standard TWFE event study, which produces “negative weights” artifacts. Fails (see did-staggered-heterogeneous-effects).

References

• Borusyak, K., Jaravel, X., & Spiess, J. (2024). Revisiting event-study designs: Robust and efficient estimation. Review of Economic Studies, 91(6), 3253–3285.
• Callaway, B., & Sant’Anna, P. H. C. (2021). Difference-in-differences with multiple time periods. Journal of Econometrics, 225(2), 200–230.
• Sun, L., & Abraham, S. (2021). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. Journal of Econometrics, 225(2), 175–199.