Policy Evaluation Panel Event Study
Policy Evaluation Panel Event Study Design · Also known as: panel event study, event-study DiD, staggered event study, difference-in-differences event study
A panel event study is a quasi-experimental design that traces how an outcome evolves in periods before and after a policy event, using unit and time fixed effects to identify the causal effect. Widely used in economics and policy research, it tests for anticipation effects, verifies parallel pre-trends, and estimates dynamic treatment effects across post-treatment horizons — making it the standard toolkit for rigorous policy evaluation with observational panel data.
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.
When to use it
Use a panel event study when you have multi-period panel data and a policy or shock that is imposed on some units but not others (or hits units at different times), and you want to estimate both immediate and dynamic causal effects. It is appropriate when the parallel-trends assumption is defensible and a credible comparison group exists. Do not use it when all units are treated simultaneously with no pre-treatment observations, when the panel is very short (fewer than 3 periods total), when treatment status is endogenous to the outcome in a way that cannot be addressed by fixed effects, or when sample sizes per cohort are very small — in those cases, synthetic control or IV designs are preferable.
Strengths & limitations
- Produces a full time profile of treatment effects, revealing anticipation, dynamic adjustment, and whether effects persist or fade.
- Pre-trend plot provides a visible, intuitive falsification test of the parallel-trends assumption.
- Unit and time fixed effects absorb both time-invariant unobserved heterogeneity and period-specific common shocks.
- Modern heterogeneity-robust estimators (Callaway-Sant'Anna, Borusyak et al.) handle staggered adoption without negative-weighting bias.
- Flexible enough to accommodate unbalanced panels and varying treatment cohort sizes.
- Parallel trends is an untestable assumption for post-treatment periods; the pre-trend test is necessary but not sufficient.
- With staggered adoption and treatment effect heterogeneity, the classical TWFE estimator produces biased and potentially sign-reversed estimates — requiring modern robust alternatives.
- Loses precision rapidly as the number of relative-period indicators grows, especially in panels with few pre- or post-treatment periods.
- Interpreting cohort-aggregated ATTs requires care: the weighted average may mask substantial heterogeneity across treatment cohorts.
- Not identified when there is no clean comparison group (e.g., universal or simultaneous treatment).
Frequently asked
What is the difference between a panel event study and a standard DiD?
A standard DiD produces a single average treatment effect (the interaction coefficient). A panel event study estimates a separate coefficient for each relative time period around the event, giving a full dynamic picture of how the effect evolves — and also providing a formal pre-trend check that a simple DiD cannot.
Why is the classical TWFE estimator problematic with staggered adoption?
When units are treated in different periods and treatment effects vary across cohorts or over time, the TWFE interaction-weighted estimator can assign negative weights to some cohort-period comparisons, yielding a biased or even sign-reversed estimate. Robust estimators (Callaway-Sant'Anna, Borusyak-Jaravel-Spiess) avoid this by constructing clean cohort-specific comparisons before aggregating.
How many pre-treatment periods do I need?
At least two pre-treatment periods are needed to plot a meaningful pre-trend and run a formal test. Three or more are strongly preferred, as a single pre-period point provides no visual trend information.
What if my pre-trend test fails?
A significant pre-trend means the comparison group was already on a different trajectory, violating parallel trends. Options include: restricting the sample to better-matched control units, conditioning on covariates, using a synthetic control as the counterfactual, or adopting a doubly robust estimator that adjusts for time-varying covariates.
Can I use this method with a binary or count outcome?
Yes, though interpretation differs. The panel event study with fixed effects is a linear probability model for binary outcomes, which can produce out-of-range predictions but remains consistent for the average partial effect. For counts, Poisson fixed-effects event studies are common. Non-linear alternatives must be checked for incidental-parameters bias in short panels.
Sources
- 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 ↗
- Borusyak, K., Jaravel, X., & Spiess, J. (2024). Revisiting event study designs: Robust and efficient estimation. Review of Economic Studies, 91(6), 3253-3285. DOI: 10.1093/restud/rdae007 ↗
How to cite this page
ScholarGate. (2026, June 3). Policy Evaluation Panel Event Study Design. ScholarGate. https://scholargate.app/en/causal-inference/policy-evaluation-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
- Instrumental Variables in Health ResearchHealth Economics↔ compare
- Panel Fixed EffectsEconometrics↔ compare
- Propensity Score MatchingResearch Statistics↔ compare
- Regression DiscontinuityCausal inference↔ compare
- Synthetic ControlCausal inference↔ compare