Measuring Policy Diffusion through the Text Similarity across Millions of Legislative Bills

Policy diffusion has been an important topic across subfields of political science, yet existing studies have focused on a few policies that are known to be diffused.  This approach may well overlook unknown policy diffusion and may lead to misleading empirical analysis due to the selection on the dependent variable.  We propose a novel measure of policy diffusion based on the text similarity for all combinatorial pairs of legislative documents.  Our method is scalable to millions of documents, and therefore it allows researchers to compute the measure across all policy areas.  We compute the proposed measure using about two million legislative bills introduced in all US state legislatures and the Congress from 1995 to 2015 and show descriptive analysis to examine existing hypotheses on policy diffusion.