Robust Theories of Political Polarization

• In Event: Grassroots Politics

Abstract

The temptation to infer a causal relationship between correlated variables has been a long- standing affliction of social science, the study of political polarization being no exception. Indeed, recent work has suggested a connection between political polarization and one of a rotating bevy of other variables, most prominently income inequality, based on observed national-level co-trends. Yet are such correlations even consistent with a potentially causal relationship? We begin by noting that polarization, as far as rational choice models go, will always be a phenomenon explained by across-constituency variation, which contributes to but is not identical to national variation – the measure most commonly used in the literature. This prompts us to ask, given a model of elections, 1) what changes in the exogenous variables of the model would need to obtain for it to predict an increase in polarization, and 2) what additional assumptions would an analyst need to make in order to draw conclusions about polarization? To address these questions, we propose a general approach based on the potential outcomes framework that puts observed polarization in direct conversation with existing models of elections. We demonstrate this approach using robust comparative statics and data-driven simulations, employing as an example throughout the hypothesis that rising income inequality has led to increases in political polarization.

Authors

• Asya Magazinnik, MIT

• Dan Alexander, University of Rochester