Article Summary
Introduction
U.S. politics increasingly revolves around group-based identities and us vs. them conflicts, especially as they relate to politics. This has led many to question how enmeshed partisan group identity is with other such identities. In their paper, Westwood and Peterson contend race and partisanship, two fundamental cleavages in American political behavior, are inescapably intertwined; race and partisanship both cue the other in the minds of voters.
Specifically, the authors build upon psychology literature on “spreading activation,” where the activation of certain concepts in the minds of voters activates other, related concepts. They pose the question of whether race and partisanship are co-activated in such a manner, such that when voters update their opinions on one group, opinions on the other are also influenced. Westwood and Peterson call this process “parallel updating.” Their research speaks to the potential bi-directionality of the race-partisanship relationship, and how affective polarization may be a function of non-partisan but racial day-to-day interactions.
Analytical Approach
The authors conduct a series of 3 studies (detailed in the table below) where respondents are asked to take part in a mock behavioral “trust” game. In each game, the authors randomize the quality of the interaction the respondent has with the fictional opposing player. Trust games are typically used to measure cooperation with out-groups, where players allocate money to each other in varying amounts. The basic structure makes it such that the more a player expects reciprocity from the other, the more money they will allocate.
| Study | Description | Treatments | Outcomes |
|---|---|---|---|
| Study 1: Field Experiment in an Online Labor Market | Workers of an online labor market are asked to copyedit a website with randomized partisan cues | Control: Text suggested founders met working for a non-profit Republican: founders met working for Republican party Democrat: founders met working for the Democratic party | Reservation wage (wage required for worker to be contracted again), number of errors caught, number of corrections made |
| Study 2: Partisan Congruity and Consumer Choice | Respondents asked whether they would like to purchase a gift card at a discounted price, with randomized partisan cues | Baseline: email suggests cards were leftover from work with a non-profit Republican: leftover from work with Republican campaigns Democrat: leftover from work with Democratic campaigns | Whether respondent responded to offer, whether responded continued to purchase card |
| Study 3: An Incentivized, Population-Based Survey Experiment | Respondents had to choose between two cash payments for their survey completion | Baseline: Choice between receiving $3 or $6 and $4 donated to opposing party national committeeHigher Payment: Choice between $3 or $9 and $4 donated to opposing partyReligion: $3 or $6 and $4 to religious outgroup (American atheists, Christian legal society) Geography: $3 or $6 and $4 to Association of Western (Eastern) States (based on where they lived) | Choice response |
The modified game proceeds in four steps. (1) The respondent learns they are assigned the role of player 2, and that player 1 can see information about them regarding their age, gender, race, income, and political party. (2) The respondent is given the same information (varied depending on study) about player 1. (3) The respondent sees the result of player 1’s move: their allocation of money toward the respondent. This component is randomized to be either a large amount of money ($8) or none. The respondents are then prompted to re-allocate some of this money (tripled) back to player 1. (4) The respondent is given player 1’s reason for their allocation (race or party).
Depending on the study, the dependent variable was either out-party/race affect (measured using feeling thermometers given post-treatment) or the allocation of money made to the other player. For the behavioral outcome, participants in study 2 repeated four additional rounds of the trust game, but this time as player 1 and with randomized race and partisanship of player 2. If “parallel updating” is occurring, the effect on race/party feelings and allocations should be statistically significant and substantively similar.
Main Findings
The authors find strong evidence of parallel updating for race and party. In Study 1 (racial treatment), having a negative experience with the other player (who was a member of the out-racial group) resulted in a 10-point drop in out-racial affect. A similar drop in affect was found for out-partisans as well, even when the partisanship of player 1 was randomized. These findings persist in the behavioral results in study 2: respondents with negative experiences with player 1 in the previous game due to race gave around $2 less to both out-racial and out-partisan group members in subsequent games. Finally, with the partisan treatment in study 3, respondents reported cooler feelings towards both out-racial and out-party members (although the effect on out-party was greater). In total, when respondents have a negative experience with a member of the out-race (party) group, they update their opinions and behavior toward both out-race (party) and out-party (race) members in a similar fashion.
The authors perform a number of robustness checks. In study 1, they also measure out-income class affect, as well as affect toward immigrants and baseball players as placebo tests. If the negative treatment is causing more generalized negative affect, it should be evident from these measures. However, while affect toward out-class members decreases, it does so only very mildly, and not at all for immigrants and baseball players. Additionally, the results persist even when respondents have access to party information (study 1), for those who have sorted vs. unsorted partisan/racial identities, are not moderated by racial resentment, and when subsetting by racial or partisan subgroup.
Implications
This paper provides direct evidence of a strong, seemingly automatic relationship between partisanship and race in the minds of the American electorate. Other research has hinted at interlinked affective evaluations of such groups, but this is the first paper to provide direct evidence of the relationship. Additionally, the authors expand upon previous literature on how race influences partisan attitudes by documenting how partisanship also influences racial attitudes. This has implications for the future of affective polarization, as negative evaluations of one group are likely to lead to negative evaluations of the other, potentially creating a self-fulfilling loop of negativity.
More broadly, the paper has additional implications for political identity. Namely, it suggests such identities are deeply related to non-political facets of life. To fully understand political identity, researchers should not limit themselves to explicitly political considerations. It is possible (and perhaps likely) that other group identities are also interrelated in a similar, if not potentially weaker manner.
Questions left unanswered
The authors give a thorough description of the relationship between group evaluations within a fairly rapid timeframe, measuring “parallel updating” immediately after a survey treatment. The paper does not answer, however, the durability of these updates. More specifically, there is an open question as to whether these affective evaluations decay in an equally parallel manner. It is possible that after more automatic association of one group to another, more active cognitive processing weakens the relationship.
Additionally, one finding from the paper (the asymmetry in parallel updating in response to the partisan treatment) goes slightly against expectations, and the authors can only hypothesize as to the reason. They posit this could be due “social norms constraining self-reports of hostile feelings towards out-race members”, but cannot confirm this mechanism with existing data (and similar forces should also be at play in the racial treatment)
Methods and Analysis
Was the study and its analyses pre-registered?: No
Did the study rely on proxy variables to measure polarization?: Yes
In Study 2, authors used a behavioral game as the dependent variable. Typical feeling thermometer measures were used in the other two studies.
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 authors concede a number of limitations to their study. First, the games are purposefully artificial in construction, so the generalizability of the purported effects in real-world scenarios may be more muted. Additionally, the analysis is limited to two racial groups: white and Black Americans. Parallel updating in affect may be different for groups with weaker associations with partisanship. Finally, the authors note there is no obvious out-group for political independents, so it is unknown how such respondents would react in the trust game setting.
Lastly, there is some ambiguity as to whether the effect of decreasing out-group affect is specific to party and race or is a more generalized decrease in out-group affect. The authors attempt to alleviate this concern with the inclusion of class attitudes, but the economic classes they randomize are limited (four groups from $30K - $70K). The placebo tests are also not specific to out-groups. This leaves open the possibility that affect would have decreased for any out-group of salience, and we see a strong connection between party and race only because those were the groups of focus.
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
Westwood, S. J., & Peterson, E. (2022). The inseparability of race and partisanship in the United States. Political Behavior, 44(3), 1125–1147. https://doi.org/10.1007/s11109-020-09648-9
Bibtex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
@article{westwood_inseparability_2022,
title = {The {Inseparability} of {Race} and {Partisanship} in the {United} {States}},
volume = {44},
issn = {1573-6687},
url = {https://doi.org/10.1007/s11109-020-09648-9},
doi = {10.1007/s11109-020-09648-9},
language = {en},
number = {3},
urldate = {2023-08-22},
journal = {Political Behavior},
author = {Westwood, Sean J. and Peterson, Erik},
month = sep,
year = {2022},
keywords = {Polarization, Affect, Social identity},
pages = {1125—1147}
}