A Dynamic Hierarchical Bayesian Measurement Model of Political Risk

J. Sumner

Abstract:Political risk has long been theorized as a key factor that drives the allocation of foreign direct investment (FDI). Frequently defined as the likelihood a host government will engage in actions that threaten an investor’s profitability, political risk is thought to be remedied by political institutions that allow countries to make more credible commitments to honor their pre-investment promises. As a latent variable, scholars have struggled with ways to link institutions to FDI flows in the absence of a measure of political risk. This paper presents a dynamic hierarchical model of political risk that aims to address this problem by producing from observable indicators a single time-series cross-sectional measure of political risk. Additionally, this model produces estimates that allow scholars to measure within-country variation in political risk, a previously unestimated quantity. The model itself is flexible enough to be of use to comparative scholars more generally to estimate latent variables at multiple levels of a hierarchy from observable indicators produced at different levels of that hierarchy. Many of the phenomena that interest political scientists are not directly measurable. Often, this leads political scientists to search for and employ proxy variables in order to test theories of substantive interest in the absence of direct measures. A good proxy variable should translate very directly from the concept of interest, and some concepts more readily lend themselves to good proxies. For instance, while we cannot directly measure political competition, we can do a fair job of proxying for it by measuring the margin of victory in an election, or while we cannot directly measure violence, we can measure the number of civilians killed in a conflict. ∗Contact: jane.lawrence@emory.edu

Economics,Political Science