Populism and Heterogeneous Effects: The Causal Mechanisms Identification Problem

Abstract

This paper argues that empirical political science has conflated two distinct identification problems: identifying treatment effects and identifying the causal mechanisms through which treatments operate. The credibility revolution strengthened the first task but was not designed to solve the second. Yet heterogeneous treatment effects are routinely interpreted as evidence of mechanisms, as though identifying a causal effect licenses inference about the pathway that produced it. The paper shows that such inferences are underdetermined when political outcomes aggregate multiple latent causal pathways into a single equilibrium category. Three empirically confounded sources of heterogeneity are as follows: scale heterogeneity, generated by nonlinear outcome transformations; threshold heterogeneity, arising from variation in proximity to decision boundaries; and mixture heterogeneity, arising from equifinal causal pathways. A latent-index framework demonstrates that significant interactions can emerge even when the structural treatment effect is homogeneous and there is no mechanism heterogeneity. An empirical illustration using the Colantone-Stanig trade-exposure data reproduces canonical interaction patterns in a constant-β simulation, showing that heterogeneous responses may arise in the absence of heterogeneous mechanisms.