Panel Event Study
Panel Data Event Study Design · Also known as: event-study regression, dynamic DiD, relative-time regression, distributed-lag panel model
A panel event study estimates the dynamic causal effect of a treatment or policy by regressing an outcome on a full set of relative-time indicators — one for each period before and after the event — while controlling for unit and time fixed effects. The resulting coefficient plot shows how the treated units diverged from untreated units at each point in calendar time relative to their treatment date, making both pre-treatment trend violations and post-treatment effect trajectories immediately visible.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
+10 more
When to use it
Use a panel event study when you have repeated observations for multiple units, each receiving a well-defined treatment at a known date (which may differ across units), and you want to estimate the full dynamic trajectory of the effect rather than a single average. It is especially valuable when the timing of treatment varies across units, providing additional identification power. It is not appropriate when treatments are ongoing with no clear start date, when the panel is very short (fewer than three pre-treatment periods make pre-trend testing uninformative), when treatment is simultaneously assigned to all units (no control group exists), or in settings with severe heterogeneous treatment timing where simple two-way fixed effects estimates can be sign-reversed (in those cases, prefer heterogeneity-robust estimators such as Callaway and Sant'Anna).
Strengths & limitations
- Reveals the full dynamic time profile of treatment effects, distinguishing immediate from delayed or fading impacts.
- Provides a built-in falsification check: pre-treatment coefficients that are jointly zero support the parallel-trends assumption.
- Exploits variation in treatment timing across units, increasing identification power relative to a single-cohort design.
- Handles unbalanced panels and missing data more flexibly than aggregate before-after comparisons.
- Produces an intuitive coefficient plot that communicates both the magnitude and the timing of effects to non-specialist audiences.
- With staggered treatment adoption, the standard two-way fixed-effects version can produce misleading estimates if treatment effects are heterogeneous across cohorts; heterogeneity-robust alternatives (e.g., Callaway and Sant'Anna) are then needed.
- Requires a sufficient number of pre-treatment periods — at least three to four — to test parallel trends meaningfully; very short panels cannot support this diagnostic.
- Inference is complicated by serial correlation within units; failing to cluster standard errors at the unit level typically overstates precision.
- Binning of distant relative-time periods (due to thin data) can mask important dynamics at the tails of the event window.
Frequently asked
How is a panel event study different from a simple event study?
A classic finance event study estimates abnormal returns around a single corporate event using only the treated security's time series, without a panel of control units. A panel event study includes a comparison group (never-treated or not-yet-treated units) and adds unit fixed effects, making it a difference-in-differences design that traces out dynamic effects and enables a pre-trend test.
Why must I omit one pre-treatment period?
The relative-time dummies for all periods, together with the unit and time fixed effects, span a collinear system. Omitting one period (conventionally ell = -1) breaks the collinearity and anchors all coefficients to that baseline period, whose implied effect is zero.
What if my pre-trend coefficients are not all near zero?
Non-zero pre-treatment coefficients indicate that treated and control units were already diverging before the event, which invalidates the parallel-trends assumption. You should revisit your choice of control group, consider conditioning on pre-treatment covariates, or use a method that does not rely on parallel trends, such as the synthetic control method.
When should I use a heterogeneity-robust estimator instead?
When treatment is staggered (units receive treatment at different calendar dates) and you suspect that treatment effects differ across adoption cohorts, the standard two-way fixed-effects panel event study can be biased. Estimators by Callaway and Sant'Anna (2021) or Sun and Abraham (2021) decompose effects by cohort and are recommended in that setting.
How many pre-treatment and post-treatment periods should I include?
Include as many pre-treatment periods as your data permit, to give the parallel-trends test sufficient power. For post-treatment periods, balance statistical precision (estimates become noisier at long horizons due to thin data) with the substantive question of how long the effect persists. Distant bins with few observations are commonly collapsed or trimmed and this choice should be reported.
Sources
- Freyaldenhoven, S., Hansen, C., Perez-Orive, J., & Shapiro, J. M. (2021). Visualization, Identification, and Estimation in the Linear Panel Event-Study Design. NBER Working Paper 29170. National Bureau of Economic Research. link ↗
- Callaway, B., & Sant'Anna, P. H. C. (2021). Difference-in-Differences with Multiple Time Periods. Journal of Econometrics, 225(2), 200-230. DOI: 10.1016/j.jeconom.2020.12.001 ↗
How to cite this page
ScholarGate. (2026, June 3). Panel Data Event Study Design. ScholarGate. https://scholargate.app/en/causal-inference/panel-event-study
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Difference-in-DifferencesEconometrics↔ compare
- Dynamic Difference-in-DifferencesCausal inference↔ compare
- Panel Fixed EffectsEconometrics↔ compare
- Synthetic Control MethodCausal inference↔ compare