Article Summary
Introduction
Two of the most central and durable findings in contemporary research in political behavior have been (1) the growing importance of partisanship as a social identity and (2) the increasing animosity between partisans. Naturally, many have linked the two (either implicitly or explicitly), suggesting partisanship as a stronger social identity increases negative affect toward the rival party and positive affect toward one’s own party. However, seemingly no research has been done attempting to isolate the effect of partisanship as a social identity on affective polarization; while partisanship or other partisan signals are manipulated, partisanship as a social identity remains constant.
In their paper, West and Iyengar attempt to both confirm the importance of partisanship as a social identity and determine its connection with affective polarization through a survey experiment administered over a panel. They contend partisan identities should be strongest in times of heightened political salience, as such social identities are sensitive to differences in context. Furthermore, as partisan identities are theoretically construed as social rather than individual identities, heightening individual identities ought to reduce the importance of social/partisan identity. They theorize this reduction in the importance of social identity may decrease negative out-party affect, as perceptions of group threat are assuaged. In implementing their study, the authors contribute to a large literature documenting both the underpinnings of affective polarization and the content of partisan identity (as both a social identity and an ideological set of positions).
Analytical Approach
The authors embed an experiment in a two-wave panel survey (Wave 1 n = 2,513, Wave 2 n = 1,311, combined n = 1,266) through Dynata during and several weeks after the 2018 midterm elections. In both waves, all respondents are asked to complete a battery of questions that form a partisan identity index, as well as feeling thermometers measuring affective polarization and perceptions of traits. The panel structure of the survey allows the authors to compare levels of partisan social identity during times of heightened partisan salience (during the midterm elections) and reduced partisan salience (during the holiday period).
In each wave of the survey, respondents are randomly assigned to one of three treatment conditions meant to either focus the respondent on themselves as individuals or on others (prior to the measurement of attitudes). These treatment conditions were assigned independently of wave or prior treatment status (so a respondent could be treated in one wave but not the other) and are listed in the table below. Respondents in the self-affirmation condition were asked to respond to how much certain values related to them, whereas the “ranking other” control condition asked them to rank applicability of certain traits to a group of strangers. The true control condition simply listed the traits, but a manipulation check revealed post-treatment evaluations of self were almost identical between the self-affirmation and pure control conditions, so they were combined by the authors.
Table 1: Treatment conditions for Study
| Treatment | Description |
|---|---|
| Self-Affirmation: Ranking Self | Respondent shown values with brief description, asked to rate themselves (1-5, Very Much Like/Unlike Me) on how each value related to them |
| Self-Affirmation: Ranking Other | Respondents shown values with brief description, asked to rate their applicability to group of strangers (picture shown) |
| Pure Control | Respondents shown values with brief description |
The authors then measure differences in both partisan self-identity and partisan affect between treatment conditions within-wave to determine if an individual focus can decouple one’s self from partisan identity and if that decoupling is further related to a decrease in negative out-party affect.
Main Findings
While partisan identity does exhibit classic traits of other social identities, the authors find their treatment inducing reductions in partisan identity does not decrease levels of affective polarization.
First, the authors do find a significant reduction in partisan self-index from wave 1 to wave 2, indicating the salience of partisan context moderates levels of identification with party. This effect is consistent and large across parties, with Democrats and Republicans showing 40% and 48% reductions in partisan self-index, respectively.
Second, the authors find evidence that a focus on individual identity also reduces partisan identity. Respondents receiving either the ranking-self or pure control conditions had slightly lower levels of partisan self-index (0.21 on a 0-12 scale). This effect is entirely attributable to Democrats, who showed an 18% of a standard deviation reduction in partisan self-index. Republicans had no significant differences between treatment conditions. In combination with the first finding, these results largely confirm the classification of partisan identity as a social identity.
However, in the final set of results, the authors find no consistent significant differences in levels of affective polarization (for both feeling thermometers and trait evaluations) between treatment conditions. While affective polarization did decrease slightly between waves, the size of this effect is a fraction of the size of the difference in partisan self-index. Within wave, disaggregating by partisanship does not change the lack of difference between treatment conditions, nor does modeling the treatment as an instrumental variable for partisan self-index.
Implications
These findings strongly suggest partisanship is an important social identity: its strength varies as a function of group salience and focus on individual versus group characteristics. As such, we should expect partisan identity to continue to be a strong influence on behavior so long as partisanship continues to be highly salient in our day-to-day lives. It also suggests there are opportunities for the social aspects of partisanship to be mitigated through a de-emphasis on group dynamics.
However, it does not appear that affective polarization is strongly tied to partisanship as a social identity. This suggests other facets of partisanship less malleable by social identity (such as ideology) may be more responsible for differences in affective polarization. This speaks to a broader literature attempting to determine whether affective polarization is rooted in affect, ideology, or party loyalty (see Orr and Huber 2020, Dias and Lelkes 2022, and Orr, Fowler, and Huber 2023). At minimum, the findings emphasize a need for further research into the (dis)connection between group identity and group affect.
Questions left unanswered
While the results are suggestive of some other dimension of partisanship being responsible for affective polarization, the authors are unable to do more than speculate as to what that dimension is. Generally speaking, it is difficult to answer questions about the effects of partisan identity without a firm grasp of what components of partisanship are related to social identity and which aren’t.
Additionally, while the treatment was meant to increase individual-level focus, it is unclear what would occur if the researchers instead focused on increasing the salience of group connections. It is possible the lack of effects the authors find is because there’s some floor to affective polarization baked into partisanship, but affective polarization can still be heightened when group attachment is elevated.
Methods and Analysis
Was the study and its analyses pre-registered?: No
Did the study rely on proxy variables to measure polarization?: No
Were standard p-value thresholds used (p<.05 or 95% Confidence Intervals that don’t overlap zero)?: Yes
- Largest p-value presented as significant: 0.05
Were correlational results interpreted with causal language?: No
Limitations / Weaknesses
The crux of this paper depends on the treatment effectively increasing focus on the self at the expense of one’s group. The finding that the treatment only reduces partisan self-index among Democrats is suggestive that the treatment is fairly weak (as the authors themselves admit). But the real concern is how durable the treatment is for the duration of the survey. The treatment is itself followed by a number of manipulation checks, then the partisan index questions, and then the affective polarization questions. Given the number of items, it would seem plausible the effect of the treatment has worn thin by the time the respondent reaches the affective polarization questions. Perhaps the wording of those questions could also revert respondents back into their pre-treatment group-thinking ways. While it is admittedly difficult to independently manipulate partisan identity, future work should consider a stronger treatment and test its durability further into the survey.
Open Data & Analyses
Does the article make the replication data publicly available?: Yes
Does the article make the replication analysis scripts publicly available?: Yes
Article Citation
West, E. A., & Iyengar, S. (2022). Partisanship as a social identity: Implications for polarization. Political Behavior, 44(2), 807–838.
Bibtex
1
2
3
4
5
6
7
8
9
10
11
@article{West_Iyengar_2022,
title={Partisanship as a social identity: Implications for polarization},
volume={44},
DOI={10.1007/s11109-020-09637-y},
number={2},
journal={Political Behavior},
author={West, Emily A. and Iyengar, Shanto},
year={2022},
month={Jun},
pages={807–838}
}