Affective Polarization, Emotions, and Participation: The Cost of Action Matters

• In Event: What Is Affective Polarization and How Should We Measure It?

Abstract

Despite the centrality of affect to the concept of affective polarization, few studies of polarization examine negative manifestations of affect other than anger (Fischer & Lelkes, 2023). This is surprising, since affective polarization has been defined by several dimensions of negative affect, such as dislike, distrust, fear, and loathing (Iyengar & Westwood, 2014; Druckman & Levendusky, 2019). These dispositions are distinct, and it remains unclear to what extent distinguishing between them matters for affective polarization research. For example, both fear and anger have been shown to correlate with affective polarization (Lu & Lee, 2019; Renström et al., 2023), but research from political psychology suggests that fear and anger drive political mobilization in opposite directions (Marcus et al., 2000; Valentino et al., 2011).

Therefore, an overfocus on anger draws into question whether affective polarization is a theoretically distinct construct or, conversely, a conceptual repackaging of partisan-directed anger as a ‘raw, reflexive emotion’ (Iyengar et al., 2019, p. 141). Moreover, a methodological overreliance on surveys draws into question whether affective polarization has demonstrable, real-world effects on political behavior outside of self-reports.

In this study, we address both questions through a pre-registered online experiment (N=351) conducted in November 2023 on US adults. We induced partisan-directed fear and anger through an autobiographical recall task and asked participants to complete standard feeling thermometers. Then, participants were endowed with a financial bonus ($2), which they could then use to influence whether a $1,000 donation goes to the Democratic of Republican Party. This allows us to test how affective polarization, anger, and fear (IVs) relate to each other in the context of an actual, real-world measure of political participation (DV). By systematically varying the costs to influence this donation across treatments, we also test to what extent the predictive value of these IVs and their unique contribution to predicting political participation is cost dependent. Furthermore, we compare this incentivized measure of political participation with self-reports obtained through traditional survey measures.

Our results contribute to the study of affective polarization in three key ways. First, we show that affective polarization – not anger – strongly predicts political participation but only when the costs to participate are low (i.e., in self-reports and low-resource actions). Second, when the costs to participate are high, anger – but not affective polarization – drives political participation. Third, the differences observed in the effects of anger and affective polarization on political participation suggest that indeed, affective polarization is a valid theoretical construct that is conceptually distinguishable from anger.

Therefore, in addition to demonstrating the concept validity of affective polarization, our findings have major implications for the external validity of affective polarization research. If our results are generalizable, then the vast bulk of extant research on affective polarization and participation may only apply to ‘cheap,’ low-cost forms of participation. When the costs to participate are high and money is on the line, anger appears to be a better predictor of participation than affective polarization.

Status of the Paper:

While the experiment is finished and the data is analyzed, we plan to run a second experiment in February 2024 using a different emotion induction method (manipulated news stories). The reasons for a second experiment are two-fold. First, we used an autobiographical recall task in the first experiment. Our pilots showed a high amount of ChatGPT-generated answers from MTurkers, and we achieved better results after incorporating some NLP checks and using MTurkers with a Master Qualification. However, we could only recruit a limited number of Master MTurkers based in the US, which is why the experiment is low-N relative to our pre-analysis plan. Second, although our decision to use autobiographical recall is well-grounded in prior research, we aim to replicate our results with an alternative emotion induction method. Therefore, at the time of applying we have a draft of the first experiment, but plan to update the paper with results of the second experiment before the conference.

Authors

• Mike Farjam, Lund University

• Michael Bossetta, Lund University, Sweden