On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data
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 ab...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
Cambridge University Press (CUP)
2021
|
Online Access: | https://hdl.handle.net/1721.1/138140.2 |
_version_ | 1811083124733378560 |
---|---|
author | Imai, Kosuke Kim, In Song |
author2 | Massachusetts Institute of Technology. Department of Political Science |
author_facet | Massachusetts Institute of Technology. Department of Political Science Imai, Kosuke Kim, In Song |
author_sort | Imai, Kosuke |
collection | MIT |
description | 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. |
first_indexed | 2024-09-23T12:23:18Z |
format | Article |
id | mit-1721.1/138140.2 |
institution | Massachusetts Institute of Technology |
language | English |
last_indexed | 2024-09-23T12:23:18Z |
publishDate | 2021 |
publisher | Cambridge University Press (CUP) |
record_format | dspace |
spelling | mit-1721.1/138140.22021-11-15T20:06:03Z On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data Imai, Kosuke Kim, In Song Massachusetts Institute of Technology. Department of Political Science 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. 2021-11-15T20:06:02Z 2021-11-15T16:52:37Z 2021-11-15T20:06:02Z 2021 2021-11-15T16:39:31Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/138140.2 Imai, K., & Kim, I. (2021). On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data. Political Analysis, 29(3), 405-415. en http://dx.doi.org/10.1017/PAN.2020.33 Political Analysis Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/octet-stream Cambridge University Press (CUP) MIT web domain |
spellingShingle | Imai, Kosuke Kim, In Song On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data |
title | On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data |
title_full | On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data |
title_fullStr | On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data |
title_full_unstemmed | On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data |
title_short | On the Use of Two-Way Fixed Effects Regression Models for Causal Inference with Panel Data |
title_sort | on the use of two way fixed effects regression models for causal inference with panel data |
url | https://hdl.handle.net/1721.1/138140.2 |
work_keys_str_mv | AT imaikosuke ontheuseoftwowayfixedeffectsregressionmodelsforcausalinferencewithpaneldata AT kiminsong ontheuseoftwowayfixedeffectsregressionmodelsforcausalinferencewithpaneldata |