Two-way Fixed Effects and Differences-in-Differences Estimators in Heterogeneous Adoption Designs
Date: Thursday, Apr 10, 2025, 16:00 ~ 17:45
Speaker: Xavier D'Haultfœuille (CREST-ENSAE)
Location: Webinar
◈ Keynote : "Two-way Fixed Effects and Differences-in-Differences Estimators in Heterogeneous Adoption Designs" by Xavier D'Haultfœuille (CREST-ENSAE)
◈ AEW Discussion : "Estimating Treatment Effects in Panel Data Without Parallel Trends," by Shoya Ishimaru
◈ 발표자 : Xavier D'Haultfœuille (CREST-ENSAE)◈ 일 시 : 2025년 4월 10일 목요일 16:00 ~ 17:45
◈ 장 소 : Webinar via WebEx
Webinar 참석을 위해 아래 링크를 통한 사전 등록이 필요합니다:
Two-way Fixed Effects and Differences-in-Differences Estimators in Heterogeneous Adoption Designs
Abstract:
We consider treatment-effect estimation under a parallel trends assumption, in designs where no unit is treated at period one, all units receive a strictly positive dose at period two, and the dose varies across units. There are therefore no true control groups in such cases. First, we develop a test of the assumption that the treatment effect is mean independent of the treatment, under which the commonly-used two-way-fixed-effects estimator is consistent. When this test is rejected or lacks power, we propose alternative estimators, robust to heterogeneous effects. If there are units with a period-two treatment arbitrarily close to zero, the robust estimator is a difference-in-difference using units with a period-two treatment below a bandwidth as controls. Without such units, we propose non-parametric bounds, and an estimator relying on a parametric specification of treatment-effect heterogeneity. We use our results to revisit Pierce and Schott (2016) and Enikolopov et al. (2011).
Abstract:
We consider treatment-effect estimation under a parallel trends assumption, in designs where no unit is treated at period one, all units receive a strictly positive dose at period two, and the dose varies across units. There are therefore no true control groups in such cases. First, we develop a test of the assumption that the treatment effect is mean independent of the treatment, under which the commonly-used two-way-fixed-effects estimator is consistent. When this test is rejected or lacks power, we propose alternative estimators, robust to heterogeneous effects. If there are units with a period-two treatment arbitrarily close to zero, the robust estimator is a difference-in-difference using units with a period-two treatment below a bandwidth as controls. Without such units, we propose non-parametric bounds, and an estimator relying on a parametric specification of treatment-effect heterogeneity. We use our results to revisit Pierce and Schott (2016) and Enikolopov et al. (2011).