Article Summary
Introduction
In the United States, the groups of people who call themselves Democrats and Republicans look increasingly different. They live in different kinds of places, observe different religions, are of different social classes, and so on. That is, partisanship has become aligned with other social identities. This trend is broadly called “social sorting” (or sometimes “partisan social sorting”). When parties are very socially sorted, just by knowing someone is a Democrat, you can make a pretty good guess about their race, religion, and so on. Research on affective polarization shows that partisans dislike members of the other party simply based on differences in partisanship. This article suggests that the alignment of social identities might further reinforce this hostility. Rather than simply disliking people because they are members of the opposite party, people may grow to dislike others because they are members of the opposite party, and members of a different race, and follow a different religion, all at the same time.
In this article, the author focused particularly on the emotional consequences of social sorting. Emotions are important in politics: they motivate voting, participation, and a range of other behaviors. In general, partisans become enthusiastic when their party stands to gain status and angry when their party’s status is threatened. Naturally, partisans who care more about their parties tend to feel stronger emotions at the prospect of gains or losses. The author argues that when other identities align with partisanship such that the status of multiple overlapping groups is threatened, people may be even more emotional. The author hypothesizes that political messages conveying threats to either a party’s status or its policy agenda would create anger among highly sorted partisans. Conversely, messages conveying threats to the other party’s status or agenda would create enthusiasm. Among other, less-sorted partisans, messages threatening their party’s status would create anger (albeit more weakly), but messages threatening a party’s policy agenda would not create emotion. Finally, because these emotions are theoretically driven primarily by aligned identities rather than strong opinions on political issues, the author hypothesized that people with extreme issue positions would respond emotionally to threats to their party’s policy agenda (albeit more weakly than those who are highly sorted), but not to threats to party status.
Analytical Approach
As an initial exploration of their hypothesis, the author began by examining nationally representative survey data from the American National Election Studies between 1980 and 2012. Each survey asked respondents to indicate whether each presidential candidate makes the respondent feel angry or proud. The author re-coded these variables to represent how much people feel angry or proud of their own party’s candidate and the other party’s candidate. Because partisan social sorting increased over this time period, people’s emotions toward the candidates should have diverged over time; that is, they should feel more proud of their own party’s candidate and more angry toward the other party’s candidate as time goes on.
This large-scale analysis does not allow for any causal claims. Although social sorting has increased between 1980 and 2012, many other things have also changed that might similarly explain intensifying emotional reactions to presidential candidates. As such, the author next conducted a survey experiment on 1,100 U.S. adults. Each respondent was asked to complete measures indicating how strongly they identify as Democrat or Republican, liberal or conservative, Black, Evangelical, Secular, and Tea Party. Based on these measures, a socially sorted Democrat (i.e., one with aligned identities) would be not only a Democrat, but also liberal, secular, and Black. Conversely, a socially sorted Republican would be conservative, evangelical, and identify with the tea party. Respondents also completed a set of measures capturing their attitudes about various policy issues, used later to ensure that strong identities (rather than strong opinions on political issues) are what drive political emotions.
After completing these initial measures, respondents were randomly assigned to one of four conditions in which they read a mock message from a political blog. In the first condition, the message threatened the respondent’s own party (e.g., for Republicans, the message argued that Democrats were going to win the 2012 election easily). The second condition was similar except that it threatened the party opposed to the respondent (e.g., for Republicans, the message argued that Republicans would easily win the election). In the third and fourth conditions, rather than threatening respondents’ party identities, the message instead threatened their issue positions (i.e., the policies they value) but not their party status by indicating that policies broadly associated with the Democratic or Republican parties would succeed without actually mentioning the parties by name.
| Experimental Conditions | Target of Threat | |
|---|---|---|
| In-party | ||
| Type of Threat | Party-based | 1 |
| Issue-based | 3 | |
After reading the messages, all respondents were asked how much the message made them feel angry, hostile, nervous, disgusted, anxious, afraid, hopeful, proud, and enthusiastic. The author combined the angry, hostile, and disgusted items to form a broad measure of anger, and combined the hopeful, proud, and enthusiastic items to form a broad measure of enthusiasm. The author then examined the level of anger and enthusiasm in each experimental condition, and whether those levels differ based on how socially sorted respondents were, how strongly partisan they were, and how extreme their issue positions were.
As a final step, the author compared the influence of these three variables by examining how the predicted values of anger in response to threatening messages vary across social sorting (controlling for issue polarization), partisan identity (controlling for issue polarization), and issue polarization (not controlling for anything else). They then conducted a parallel analysis focusing on enthusiasm.
Main Findings
Both partisan identity and issue polarization alone can make people respond more emotionally to messages that threaten or support the parties and their policy agendas. However, these effects are inconsistent. Strong partisans are more enthusiastic than weak partisans about messages that support their party and its policies, but they are no angrier about messages that threaten the party and policies. Similarly, people who have extreme issue positions respond with more anger to messages that threaten their own party, threaten their party’s policies, and respond with more enthusiasm to messages that support their party’s policies—but not to messages that support the party itself.
Social sorting, however, is a much stronger and more consistent predictor of emotional responses. Compared to people who are only weakly sorted, people who are highly socially sorted consistently respond with more anger and enthusiasm to threats and support of their party and its policies. However, the differences are clearest at the low end of the social sorting scale. People who are not at all sorted and have cross-cutting identities are much less angry in response to threats than the weakest partisans or people with the most moderate issue positions. In contrast, people who are highly socially sorted are very similarly angry and enthusiastic compared to strong partisans and people with extreme issue positions.
Implications
This article’s primary contribution is to show that polarization should not be reduced to partisanship or ideology alone. Most people identify with many groups. When these group identities align with partisanship, people seem particularly sensitive to threats to their party and particularly enthusiastic about support for it. When these group identities do not align with partisanship, however, people appear to take more measured responses. Rather than feeling intense emotions, they seem unbothered by threats and unenthused by support. Thus, as the parties become more socially sorted, there are fewer and fewer people with cross-cutting identities that may reduce emotional reactivity. As the author notes, however, the democratic implications of these findings are unclear. Both anger and enthusiasm can be normatively good or normatively bad, depending on context. Future work is necessary to determine whether these sorts of emotional responses—and the social sorting that seems to amplify them—are beneficial or harmful to democracy.
Questions left unanswered
The author notes that there was actually a fifth experimental condition that received no message and no emotion measures. Without this group, it is difficult to identify respondents’ baseline emotional states and, accordingly, whether the threatening message increased anger and reduced enthusiasm, whether the supportive message decreased anger and increased enthusiasm, or both. Further work might benefit from more thoroughly exploring the direction of effects.
Methods and Analysis
Was the study and its analyses pre-registered?: No
Did the study rely on proxy variables to measure polarization?: Yes
The author focuses explicitly on measures of emotion (particularly anger and enthusiasm). The author does not actually describe these variables as affective polarization in the paper, but the title implies that the author measured affective polarization even though the manuscript does not.
Were standard p-value thresholds used (p<.05 or 95% Confidence Intervals that don’t overlap zero)?: No
- Largest p-value presented as significant: 0.05
Were correlational results interpreted with causal language?: No
Limitations / Weaknesses
The author hypothesizes that weakly sorted partisans shouldn’t respond to issue-based threats, only to party-based threats. However, weakly sorted partisans do not necessarily have weak partisan identities or weak ideological identities, nor are they necessarily ideologically moderate. It is unclear from the author’s argument why social sorting should condition the effects of party-based threats more strongly than issue-based threats. This is particularly difficult because the experimental analyses of social sorting never fully control for both partisan identity and ideological extremity. Without racing the three potential moderators against each other, it is difficult to assess their relative influence on responses to issue-based and party-based threats.
This article set out to test a set of complex statistical interactions involving different types of threat, different targets of those threats, and individual level moderators (sorting, partisan identity, and issue polarization). Tests of these interactions are not formally reported (although the supplementary materials report partial tests). Testing interactions requires very large sample sizes and/or very large effect sizes. Without one or both of these, results may be prone to statistical error. The author cannot be faulted for not discussing this concern because social scientists were only just beginning to discuss it at the time the article was published. The results should be interpreted with some caution: even if threatening messages elicit anger on average, claims about who feels this anger most strongly should be somewhat circumspect (and the same is true for claims about differences in responsiveness to supportive messages).
Open Data & Analyses
Does the article make the replication data publicly available?: No
Does the article make the replication analysis scripts publicly available?: No
Article Citation
Mason, L. (2016). A cross-cutting calm: How social sorting drives affective polarization. Public Opinion Quarterly, 80(S1), 351–377. https://doi.org/10.1093/poq/nfw001
Bibtex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
@article{10.1093/poq/nfw001,
author = {Mason, Lilliana},
title = "{A Cross-Cutting Calm: How Social Sorting Drives Affective Polarization}",
journal = {Public Opinion Quarterly},
volume = {80},
number = {S1},
pages = {351-377},
year = {2016},
month = {03},
doi = {10.1093/poq/nfw001},
url = {https://doi.org/10.1093/poq/nfw001},
eprint = {https://academic.oup.com/poq/article-pdf/80/S1/351/14085340/nfw001\_suppl\_poq-15-0023-File009.pdf},
}