Treatment Effect Heterogeneity
Treatment effects that vary across units, groups, or over time (rather than a single constant effect). Heterogeneity is the wedge that breaks naive two-way fixed-effects DiD under staggered timing — when effects vary, comparisons that use already-treated units as controls put Negative-Weighting on some treatment effects. It also makes the estimand matter: different designs recover different weighted averages (e.g. IV’s LATE is a complier-weighted effect).
Relied on by
The staggered-DiD critique and heterogeneity-robust estimators; the IV/LATE interpretation.
Referenced by
- GoodmanBacon2021-DiDVariationInTiming
- SunAbraham2021-EventStudies
- DeChaisemartinDHaultfoeuille2023-TWFESurvey
- DeChaisemartinDHaultfoeuille2020-IntertemporalTE
- DeChaisemartinDHaultfoeuille2018-FuzzyDiD
- BakerLarckerWang2022-HowMuchTrustStaggeredDiD
- CallawayGoodmanBaconSantAnna-ContinuousTreatment
- Wooldridge2021-TWFEMundlakDiD
- AtheyImbens2022-DesignBasedDiD
- RothSantAnna2023-WhatsTrendingInDiD
- BakerEtAl2025-DiDPractitionerGuide
- ArkhangelskyImbens2024-FixedEffectsGeneralizedMundlak
- BorusyakEtAl2024-RevisitingEventStudyDesigns (unrestricted heterogeneity is the leading case; imposing implicit homogeneity is what breaks TWFE event studies)