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

We investigate the impact of import exposure on labor markets and voters’ support for populist parties using German data from 1987–2009. Using a standard IV model, we find that regions that faced more import exposure experienced significant labor market turmoil and increasing support for extreme-right parties. A natural question that follows is what share of the total effect of trade on politics is due to its impact on labor markets. We propose a new framework for mediation analysis in IV settings that defines the causal relations among the unobserved variables of our model and allows us to answer our question. We find that the effect of trade on political polarization is entirely explained by its impact on labor markets. Our findings are corroborated by additional analysis using distinct data sets. Our theoretical framework is general and enables identification of previously unidentified causal effects in a broad range of empirical questions that examine causal mechanisms in IV settings.