On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data
説明
<jats:title>Abstract</jats:title><jats:p>The two-way linear fixed effects regression (<jats:sans-serif>2FE</jats:sans-serif>) has become a default method for estimating causal effects from panel data. Many applied researchers use the <jats:sans-serif>2FE</jats:sans-serif> estimator to adjust for unobserved unit-specific and time-specific confounders at the same time. Unfortunately, we demonstrate that the ability of the <jats:sans-serif>2FE</jats:sans-serif> model to simultaneously adjust for these two types of unobserved confounders critically relies upon the assumption of linear additive effects. Another common justification for the use of the <jats:sans-serif>2FE</jats:sans-serif> estimator is based on its equivalence to the difference-in-differences estimator under the simplest setting with two groups and two time periods. We show that this equivalence does not hold under more general settings commonly encountered in applied research. Instead, we prove that the multi-period difference-in-differences estimator is equivalent to the weighted <jats:sans-serif>2FE</jats:sans-serif> estimator with some observations having negative weights. These analytical results imply that in contrast to the popular belief, the <jats:sans-serif>2FE</jats:sans-serif> estimator does not represent a design-based, nonparametric estimation strategy for causal inference. Instead, its validity fundamentally rests on the modeling assumptions.</jats:p>
収録刊行物
-
- Political Analysis
-
Political Analysis 29 (3), 405-415, 2020-11-12
Cambridge University Press (CUP)
