2025 Theses Doctoral

Statistical Methods for Measuring Public Opinion Polarization and Congressional Social Ties

Lee, Hane

Political polarization is a central phenomenon of contemporary American politics, yet statistical measures of polarization often remain ad hoc. This dissertation develops more rigorous statistical methods to measure polarization and relevant political behaviors. Drawing from political and economic literature, I first formalize distributional characteristics of bipolarization into axioms, mainly increased spread and bi-clustering, and introduce an index based on the Wasserstein distance that satisfies these axioms. Next, I develop two additional Wasserstein-based measures for partisan gap and issue progressivity in public opinion.

These three Wasserstein measures are applied to Ammerican public opinion data, revealing that Democratic voters are leading a growing partisan divide with a pronounced progressive shift. Lastly, I propose a fused latent factor and network model that estimates social ties among legislators directly from roll call votes. This model integrates partisan-ideological and social approaches to understanding roll call behavior. Applied to the 101st Senate, the model reveals that geographic proximity has stronger associations with social ties than conventional proxies of social connections, such as cosponsorship or shared committee membership networks.

These contributions highlights the importance of careful operationalization and methodological developments in the study of polarization and, more generally, in the measurement of political phenomena.