On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data

By Kosuke Imai, In Song Kim. Political Analysis, 2020

566 citations recorded by the source corpus.

Abstract The two-way linear fixed effects regression (2FE) has become a default method for estimating causal effects from panel data. Many applied researchers use the 2FE estimator to adjust for unobserved unit-specific and time-specific confounders at the same time. Unfortunately, we demonstrate that the ability of the 2FE 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 2FE 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 2FE estimator with some observations having negative weights. These analytical results imply that in contrast to the popular belief, the 2FE estimator does not represent a design-based, nonparametric estimation strategy for causal inference. Instead, its validity fundamentally rests on the modeling assumptions.

Subfields: Economics and Econometrics, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/660ac5f400de68ad64a9f31370bc59a30ec98d22

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Statistics and Probability

Citations in the corpus

Panel experiments and dynamic causal effects: A finite population perspective (2020, background): “For example, Sobel (2006), Athey and Imbens (2018), and Imai and Kim (2019a) study β̂TWFE in the causal literature on panel data models.”
How Much Should We Trust Staggered Difference-In-Differences Estimates? (2021, background): “…where a treatment occurs across units at different points in time, 5The body of work examining these issues include: Athey and Imbens (2018), Borusyak and Jaravel (2017),Callaway and Sant’Anna (2020), Goodman-Bacon (2019), Imai and Kim (2020), Strezhnev (2018), and Sun and Abraham (2020).”
Two-Way Fixed Effects Estimators with Heterogeneous Treatment Effects (2018, unknown): “Second, DID M is related to the multiperiod DID estimator in Imai and Kim ( 2018 ) .”
Can Technology Solve the Principal-Agent Problem? Evidence from China’s War on Air Pollution (2020, unknown): “More discussions can be found in de Chaisemartin and D’HaultfŒuille (2018, 2019); Goodman-Bacon (2018); Imai and Kim (2018). significant post-automation discontinuity of 3.0 ppb or 6.0% in NO2 concentrations and a statistically insignificant effect on reported SO2 concentrations.”
Two-Way Fixed Effects and Differences-in-Differences with Heterogeneous Treatment Effects: A Survey (2021, methodology): “The multi-period DID estimator in Imai and Kim (2021) is related to the DID M estimator.”