Measuring Partisan Competitiveness in State Legislatures: A Simulation Approach

• In Event: Measurement of Segregation and Partisanship

Abstract

Despite over half a century of academic interest in the role of interparty competition in state politics, scholars have yet to develop a commonly-agreed-upon measure of partisan competitiveness in state legislatures. Further, the most commonly used existing measures do not sufficiently account for the political geography of vote allocation among districts, an increasingly-important factor given the expanded prevalence and sophistication of gerrymandering. In this paper, we introduce an alternative approach to measuring interparty competition that explicitly models these political-geographic features using a series of Monte Carlo simulations of election results. The procedure involves two steps. First, we fit a Bayesian hierarchical model predicting two-party vote share at the district level as a function of latent district and state partisanship, as well as state and national election-specific shocks. Second, we hold the estimated partisanship fixed but draw new state and national shocks from the posterior distribution of each variable. Repeating this procedure many times gives us an estimate of how frequently control of a chamber alternates between the two parties, providing an intuitive measure of the competitiveness of the chamber. We evaluate the procedure using leave-one-out cross-validation and compare the performance of our measure to other types of commonly-used measures in the literature. Ultimately, we produce competition scores for 98 state legislative chambers over a period of 40 years.

Authors

• Wonjoon Hwang, Princeton University

• Michael Kistner, Princeton University

• William Smith, Princeton University