Analyzing Causal Mechanisms: A Review of Theory and Practices in Political Science

Abstract

Recent empirical works in political science have been showing increasing interest in the study of causal mechanisms. Given that, this paper has three main goals. First, it provides a detailed survey of the methodological approaches used by scholars to study causal mechanisms. Then, it holds practice against theory to underline the discrepancy between the methodological requirements of the most commonly used approaches and how they are implemented. Second, I point out an overlooked issue with the study of causal mechanisms; measurement error. Through a set of analytical results and Monte Carlo studies, I underline the threat of measurement error in the study of causal mechanisms. Finally, I build on the theoretical discussions to suggest five-point criteria and two novel robustness checks to facilitate the choice of the methodological approach that best matches the data structure and the theoretical framework of a given research.

Keywords: mediation analysis, causal inference, political methodology, measurement error