Flexible Causal Inference for Political Science

摘要

Measuring the causal impact of state behavior on outcomes is one of the biggest methodologicaludchallenges in the field of political science, for two reasons: behavior is generally endogenous, andudthe threat of unobserved variables that confound the relationship between behavior and outcomesudis pervasive. Matching methods, widely considered to be the state of the art in causal inference inudpolitical science, are generally ill-suited to inference in the presence of unobserved confounders.udHeckman-style multiple-equation models offer a solution to this problem; however, they rely onudfunctional form assumptions that can produce substantial bias in estimates of average treatmentudeffects. We describe a category of models, flexible simultaneous likelihood models, that accountudfor both features of the data while avoiding reliance on rigid functional form assumptions. We thenudassess these models’ performance in a series of neutral simulations, in which they produce substantialud(55% to >90%) reduction in bias relative to competing models. Finally, we demonstrate theirudutility in a reanalysis of Simmons’ (2000) classic study of the impact of Article VIII commitmentudon compliance with the IMF’s currency-restriction regime.