MerQuR+Statistics Glossary

Statistics Glossary → Event study design

Event study designconcept

An extension of difference-in-differences that estimates the effect of an intervention separately for each period before and after it, showing how the effect evolves over time and revealing any pre-trends.

Classical difference-in-differences collapses time into a before and an after period and gives a single average effect. In an event study, time is split into periods relative to the intervention (three years before, two years before, ..., the year of the intervention, one year after, ...) and a separate coefficient is estimated for each relative period. One coefficient, usually the period just before the intervention (-1), is fixed at 0 as the reference, and the others show the difference relative to it.

If the coefficients for the pre-intervention periods are close to zero, this supports the parallel-trends assumption; if they differ systematically from zero (pre-trends), the groups were already diverging before the intervention and the results are doubtful. The post-intervention coefficients show whether the effect appears immediately, grows over time or fades. In staggered designs, where units are treated at different times, the classical two-way fixed-effects estimator can mislead, and estimators developed for this problem are recommended.

An everyday example

Several cities build cycle-lane networks in different years, and a researcher studies the effect on traffic accidents. The event-study plot shows the accident gap close to zero in the years before the lanes were built (a good sign), a small drop in the first year after opening and a larger drop by the third year. This suggests the effect grows as people get used to the new lanes.

In MerQur

In Difference-in-Differences, the 'Reference relative period' option sets the period whose event-study coefficient is fixed at 0; the pre-trend test covers the periods before it.

Analyses where this concept appears

Difference-in-Differences (DiD)

Related concepts

Parallel trends assumptionTwo-way fixed effects (TWFE)