Inferring Policy Diffusion Networks in the American States

For decades scholars of state politics have studied the ways in which innovations in public policy diffuse across the states. Several studies indicate that policy diffusion is an explicitly dyadic process whereby states learn and adopt policies from their neighbors in geographic, social, economic, and political space. This dyadic diffusion process implies the existence of a policy diffusion network among the states. Using a dataset consisting of 189 policies, we introduce and apply algorithms designed to directly infer a diffusion network from a sample of policy adoption sequences. In addition to presenting the network inference algorithm, we offer three substantive contributions with regard to research on policy diffusion in the American states. First, we summarize and analyze the structure of the inferred diffusion network and assess the ways in which it has changed over the last several decades. Second, we demonstrate how the inferred diffusion network can be integrated into conventional statistical models of state policy adoption. Third, we estimate models to explain the pattern of diffusion ties and test a variety of theoretical expectations about who states choose to emulate.