Policy Evaluation Event Study Design
Policy Evaluation Event Study Design for Causal Inference · Also known as: event study, event-study DiD, dynamic DiD, PEESD
A policy evaluation event study design is a quasi-experimental approach that estimates causal effects of a policy by plotting treatment-period-by-period coefficients around a common event time. It extends difference-in-differences to visualize both pre-treatment parallel trends and the dynamic post-treatment evolution of the policy effect, and has become the standard credibility check in applied policy research.
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 policy evaluation event study design when you have panel data or repeated cross-sections where units adopt a treatment or policy at potentially different times and you want to (a) verify the parallel-trends assumption visually, and (b) characterize how the policy effect evolves over time. It is the preferred specification whenever DiD is used and multiple pre- or post-periods are available. Do not use it when all units adopt the policy simultaneously (a single treatment wave with no variation in timing), when data cover only one or two periods (too few points to plot a credible event path), or when the number of treated units is very small (fewer than 10), as event-time estimates become imprecise.
Strengths & limitations
- Provides a visible credibility check on the parallel-trends assumption through pre-treatment coefficient plots, making causal identification transparent and auditable.
- Reveals the full dynamic trajectory of a policy effect rather than a single average, which is essential when effects take time to materialize or decay.
- Accommodates staggered adoption: units that adopt at different times all contribute to identification, increasing statistical power.
- Compatible with modern heterogeneity-robust estimators (Callaway & Sant'Anna 2021) that correct biases in staggered settings with heterogeneous treatment effects.
- Widely accepted as the credibility standard in top economics and policy journals, making results directly legible to expert audiences.
- Requires enough pre- and post-treatment periods to plot a meaningful event path; with very short panels the design degenerates to a standard DiD.
- In staggered settings, classical two-way fixed-effects event study estimators can produce misleading estimates when treatment effects are heterogeneous across cohorts — the Callaway-Sant'Anna or Sun-Abraham correction should be used.
- Anticipation effects (where units change behavior before the official treatment date) cause pre-treatment coefficients to deviate from zero, complicating interpretation.
- Does not solve selection into treatment: if treated units are systematically different from controls in unobserved ways that also change over time, the parallel-trends assumption fails even with flat pre-trends.
Frequently asked
What makes this different from a standard DiD?
A standard DiD collapses the comparison to a single before-after coefficient. An event study design instead estimates a separate coefficient for every period relative to the treatment date, revealing dynamic effects and providing a visual pre-trend test that a single DiD coefficient cannot offer.
Why must tau = -1 be dropped as the reference period?
Including all event-time dummies creates perfect multicollinearity with the unit and time fixed effects. Omitting one period — conventionally the last pre-treatment period — sets the baseline and makes all coefficients interpretable as deviations from that reference point.
What should I do if I find significant pre-trends?
Significant pre-treatment coefficients indicate the parallel-trends assumption is violated. Options include controlling for unit-specific trends, using the Freyaldenhoven et al. (2019) pre-trend correction, switching to a synthetic control approach, or restricting the sample to more comparable units via matching before running the event study.
Is the classical two-way fixed-effects estimator valid with staggered treatment adoption?
Not always. When treatment effects differ across cohorts or over time, the classical estimator can produce biased and even sign-reversed estimates. Use the Callaway-Sant'Anna (2021) or Sun-Abraham (2021) estimator, which aggregate cohort-specific average treatment effects without contamination from other cohorts.
How many pre- and post-periods do I need?
There is no universal minimum, but at least three to four pre-treatment periods are recommended to meaningfully test for pre-trends. Fewer than two pre-periods provides almost no diagnostic power. Post-treatment periods depend on the policy question, but including too many can reduce precision as the number of treated observations at long horizons shrinks.
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 ↗
- Freyaldenhoven, S., Hansen, C., & Shapiro, J. M. (2019). Pre-event trends in the panel event-study design. American Economic Review, 109(9), 3307-3338. DOI: 10.1257/aer.20180609 ↗
How to cite this page
ScholarGate. (2026, June 3). Policy Evaluation Event Study Design for Causal Inference. ScholarGate. https://scholargate.app/en/causal-inference/policy-evaluation-event-study-design
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
- Interrupted Time SeriesCausal inference↔ compare
- Panel Fixed EffectsEconometrics↔ compare
- Regression DiscontinuityCausal inference↔ compare
- Synthetic ControlCausal inference↔ compare