Causal Inference for Policy Diffusion

• In Event: Causal Inference

Abstract

In all subfields of political science, researchers have long been interested in how policies diffuse across cities, states, and countries. Despite the significant progress made in the policy diffusion literature, the issue of causal inference has received little attention. Yet, this type of spatial and network analysis is precisely the area where omitted variable bias is known to be common and large. The study of policy diffusion is too important to accept large omitted variable bias. In this article, I make several contributions to improve causal analysis of policy diffusion. First, I present assumptions required to estimate causal diffusion effects and clarify why omitted variable bias is more common in diffusion studies. Second, to address this long-standing causal inference challenge, I develop a placebo test to detect a wide class of omitted variable biases using causal directed acyclic graphs. Third, I propose a difference-in-differences style estimator that can directly correct biases under an additional assumption. Finally, I provide a method for assessing the sensitivity of conclusions to potential violations of a key assumption required for the bias correction. The proposed approach is illustrated using a diffusion study of criminal justice policies. I also provide an open-source software package for the proposed methodology.

Author

• Naoki Egami, Columbia University