Methods of causal inference in contemporary political science

Abstract

This paper serves as an exposition of the causal inference methods that are most popular in political science. Rather than focusing on technical details we present a brief summary of main ideas behind each method with the goal of making them accessible to a broad audience of researchers. We also provide a research design algorithm for each method. First, we focus on a general motivation behind causal inference methods. We discuss how the problem of causality arises in hypothesis testing and describe the relationship between democracy and economic development as a case in point. Second, we give an exposition of a general causality problem within the framework of Rubin Causal Model (RCM). We provide all basic definitions and then demonstrate how the problem of causal inference arise within RCM. Third, we describe the most frequently used methods of causal inference such as randomized experiments, regression discontinuity design, difference-in-difference design, and instrumental variables. For each method we give a reader a general description as well as steps of a research design. We also briefly discuss advantages and disadvantages of each method. Armed with this knowledge, a reader can use it to find the method that is the most appropriate for a research problem at hand. We conclude by arguing that the ideas of causal inference are useful for both quantitative and qualitative research.