Instrumental Variables and Causal Mechanisms: Unpacking the Effect of Trade on Workers and Voters

PWP-CCPR-2017-002

Abstract

It is often the case that an endogenous treatment variable causally affects an intermediate variable that in turn causally affects a final outcome. Using an Instrumental Variable (IV) identifies the causal effect of the endogenous treatment on both the intermediate and the final outcome variable, but not the extent to which the intermediate variable affects the final outcome. We present a new testable framework in which a single IV suffices to also estimate the causal effect of the intermediate variable on the final outcome. We use this framework to investigate to what extent German voters responded to the labor market turmoil caused by increasing trade with low-wage manufacturing countries. We first establish that import competition increased voters’ support for only extreme (right) parties. We then decompose this populist ‘total effect’ into a ‘mediated effect’ running through labor market adjustments and a ‘direct effect’ of trade exposure on voting behavior. We find the total consists of a large populist effect driven by labor markets and a relatively smaller but moderating direct effect. Our approach provides a template that may be useful in a broad range of empirical applications studying causal mechanisms in observational data.