Taking Uncertainty Seriously: Bayesian Marginal Structural Models for Causal Inference in Political Science

Andrew Heiss and A. Jordan Nafa

Abstract

The past two decades have been characterized by considerable progress in developing approaches to causal inference in situations where true experimental manipulation is either impractical or impossible. With few exceptions, however, commonly employed techniques in political science have developed largely within a frequentist framework (i.e., Blackwell and Glynn 2018; Imai and Kim 2019; Torres 2020). In this article, we argue that common approaches rest fundamentally upon assumptions that are difficult to defend in many areas of political research and highlight the benefits of quantifying uncertainty in the estimation of causal effects (Gill 1999; Gill and Heuberger 2020; Schrodt 2014; Western and Jackman 1994). Extending the approach to causal inference for cross-sectional time series and panel data under selection on observables introduced by Blackwell and Glynn (2018), we develop a two-step pseudo-Bayesian method for estimating marginal structural models. We demonstrate our proposed procedure in the context linear mixed effects models via a simulation study and two empirical examples. Finally, we provide flexible open-source software implementing the proposed method.

Licenses

Text and figures: All prose and images are licensed under Creative Commons (CC-BY-4.0).

Code: All code is licensed under the BSD 2-Clause License.

Contributions and Code of Conduct

We welcome contributions from everyone. Before you get started, please see our contributor guidelines. Please note that this project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

References