When Should We Use Unit Fixed Effects Regression Models for Causal Inference with Longitudinal Data?

Many researchers use unit fixed effects regression models as their default methods for causal inference with longitudinal data. We show that the ability of these models to adjust for unobserved time-invariant confounders comes at the expense of dynamic causal relationships, which are permitted under...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:American journal of political science Jg. 63; H. 2; S. 467 - 490
Hauptverfasser: Imai, Kosuke, Kim, In Song
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Oxford Wiley Subscription Services, Inc 01.04.2019
Blackwell Publishing Ltd
Schlagworte:
AJPS WORKSHOP
Causal models
Causality
Data analysis
Economic models
Estimation
GATT
Identification
Inference
Membership
Regression analysis
Treatment methods
ISSN:0092-5853, 1540-5907
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Many researchers use unit fixed effects regression models as their default methods for causal inference with longitudinal data. We show that the ability of these models to adjust for unobserved time-invariant confounders comes at the expense of dynamic causal relationships, which are permitted under an alternative selection-on-observables approach. Using the nonparametric directed acyclic graph, we highlight two key causal identification assumptions of unit fixed effects models: Past treatments do not directly influence current outcome, and past outcomes do not affect current treatment. Furthermore, we introduce a new nonparametric matching framework that elucidates how various unit fixed effects models implicitly compare treated and control observations to draw causal inference. By establishing the equivalence between matching and weighted unit fixed effects estimators, this framework enables a diverse set of identification strategies to adjust for unobservables in the absence of dynamic causal relationships between treatment and outcome variables. We illustrate the proposed methodology through its application to the estimation of GATT membership effects on dyadic trade volume.
Bibliographie:The methods described in this article can be implemented via the open‐source statistical software
available through the Comprehensive R Archive Network
The initial draft of this article was entitled “On the Use of Linear Fixed Effects Regression Estimators for Causal Inference” (July 2011). We thank Alberto Abadie, Mike Bailey, Neal Beck, Matias Cattaneo, Naoki Egami, Erin Hartman, Danny Hidalgo, Rocio Titiunik, Yuki Shiraito, and Teppei Yamamoto for helpful comments.
wfe: Weighted Linear Fixed Effects Estimators for Causal Inference
http://cran.r‐project.org/package=wfe
ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0092-5853
1540-5907
DOI:10.1111/ajps.12417