(iPoster) A New Measure of Country-Level Affective Polarization

• In Event: Poster Sessions: Public Opinion 1 & Diverse Perspectives on South Asian Politics
  In Poster Session: Poster Session: Public Opinion 1

Abstract

Affective polarization refers to the widening psychological distance between different partisan groups, characterized by strong positive feelings toward one’s own party and negative feelings toward others. While the concept has been studied extensively in the United States, growing attention has been paid to affective polarization in multi-party systems, leading to the development of several measures based on like–dislike scores.

However, existing measures have notable limitations. The Reiljan index assumes that individuals support only one party and excludes non-partisans from the analysis. The Wagner index addresses these issues by measuring within-individual variation in party evaluations. When applied at the country level, however, it can yield high values even in contexts where voters largely share the same preferences, because it relies solely on within-individual variance in like–dislike scores. We argue that such situations should not be regarded as affectively polarized, as they lack conflict between voters.

We contend that affective polarization should be observed only in situations where some voters like certain parties while disliking others, and the remaining voters hold incompatible preference patterns. Such situations are characterized by both high within-individual and high within-party variance in like–dislike scores, indicating that both components are necessary conditions for affective polarization. Conceptually, this implies that the essence of affective polarization lies in the interaction between individuals and parties.

To capture this interaction, we propose the Interaction-based Polarization Score (IPS). The IPS corresponds to the squared weighted mean of the interaction sum of squares in a two-way analysis of variance without replication. It is particularly well-suited to multi-party systems, as it accommodates coalition politics, includes non-partisan voters, and simultaneously captures within-individual and within-party variation.

We apply the IPS to data from the Comparative Study of Electoral Systems covering 57 countries. Although the IPS is highly correlated with the Wagner index overall, the two measures diverge in several substantively important cases, including Germany. In the 2017 German data, the Wagner index suggests relatively high polarization, ranking Germany 8th among 39 studies in Module 5, whereas the IPS indicates weak polarization, ranking it 31st. Closer inspection shows that while most respondents expressed favorable evaluations of the two major parties, they strongly disliked the AfD. Using longitudinal polling data, the Wagner index exhibits a sharp increase in polarization following the AfD’s emergence, driven by this pattern. By contrast, the IPS remains stable, reflecting the fact that most voters hold broadly similar evaluations. We further show that the Reiljan index tends to overstate polarization in countries with large shares of non-partisans and to understate polarization in contexts where a non-negligible proportion of voters prefer more than one party.

Overall, the IPS provides a more accurate and comprehensive measure of affective polarization in multi-party systems.

Authors

• Hirofumi Miwa, Gakushuin University

• Ikuma Ogura, Hitotsubashi University

• Yuko Kasuya, Waseda University