Heterogeneous Treatment Effect Panel Event Study
Also known as: HTE panel event study, heterogeneous effects event study, staggered panel event study, CATT event study
A heterogeneous treatment effect panel event study estimates how treatment impacts vary across units and over time in a panel setting, allowing each cohort of treated units to have its own dynamic response. Seminal contributions by Sun and Abraham (2021) and Callaway and Sant'Anna (2021) showed that standard two-way fixed-effects event studies mask sign-reversing treatment heterogeneity across cohorts, motivating cohort-specific estimation followed by flexible aggregation.
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 this method when you have panel data with staggered treatment adoption across units or cohorts, and you suspect or want to test whether treatment effects differ in magnitude or sign across cohorts or over event time. It is the recommended replacement for two-way fixed-effects event studies whenever treatment timing varies across units. It is not appropriate for pure cross-sectional data, a single treated cohort with no variation in timing (a plain DiD or event study suffices), or settings where the parallel-trends assumption is fundamentally untenable. It requires enough units per cohort to estimate cohort-specific effects reliably.
Strengths & limitations
- Eliminates contamination bias that arises when already-treated units act as controls in two-way fixed-effects regressions with staggered adoption.
- Produces interpretable cohort-specific and event-time treatment paths instead of a single pooled average that can reverse sign.
- Allows flexible aggregation of cohort-period estimates into economically meaningful summaries.
- Pre-trend tests are sharper because they use only clean (never- or not-yet-treated) comparison units.
- Directly quantifies treatment-effect heterogeneity across adoption cohorts, informing policy targeting.
- Requires sufficient observations within each adoption cohort; sparse cohorts yield imprecise CATT estimates.
- Still relies on a parallel-trends assumption, now stated cohort-by-cohort; if cohort-specific trends differ from the control group, the estimates are biased.
- More complex to implement and explain than a standard two-way fixed-effects event study; software (R: did, Stata: csdid) is needed for routine application.
- When treatment timing is determined endogenously by unit characteristics that also affect outcomes, the clean comparison group assumption is violated.
Frequently asked
How does this differ from a standard two-way fixed-effects event study?
A standard TWFE event study absorbs cohort differences into unit fixed effects and implicitly uses already-treated units as controls for later-treated units. When treatment effects are heterogeneous across cohorts, the resulting estimates are weighted averages with potentially negative weights, which can produce severely biased or even sign-reversed coefficients. The heterogeneous treatment effect panel event study estimates cohort-period CATTs separately using only clean controls, then aggregates them with non-negative weights tied to a clear estimand.
What is a 'clean control group' in this context?
Units that have not yet received treatment by the period in question — either never-treated units or not-yet-treated units. Using these as the comparison group prevents already-treated units from contaminating the counterfactual, which is the key source of bias in naive staggered DiD designs.
How many pre-treatment periods do I need?
At least two pre-treatment periods are needed to conduct a credible pre-trend test. More pre-periods allow a more informative assessment of the parallel-trends assumption across cohorts. Post-treatment periods should be sufficient to trace the full dynamic path of effects.
Which software can estimate this?
In R, the did package by Callaway and Sant'Anna implements the cohort-period CATT framework directly. In Stata, the csdid command provides equivalent functionality. The eventstudyinteract package in Stata implements the Sun-Abraham interaction-weighted estimator.
Can I still use this when I have only one treatment wave?
If all treated units adopt at exactly the same time, there is only one cohort and no staggered variation to worry about. A standard two-period DiD or simple event study is sufficient, and the heterogeneous treatment effect panel event study reduces to those simpler designs. The added complexity is only justified with genuine staggered adoption.
Sources
- Sun, L., & Abraham, S. (2021). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. Journal of Econometrics, 225(2), 175-199. DOI: 10.1016/j.jeconom.2020.09.006 ↗
- 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). Heterogeneous Treatment Effect Panel Event Study. ScholarGate. https://scholargate.app/en/causal-inference/heterogeneous-treatment-effect-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
- Heterogeneous Treatment Effect Difference-in-DifferencesCausal inference↔ compare
- Panel Event StudyCausal inference↔ compare