Article Summary
Introduction
Affective polarization is on the rise among American partisans; Democrats and Republicans hold increasingly negative views of each other, disagree with each other, and seemingly avoid one another. Can anything be done to ameliorate such animosity? In their book, Levendusky and Stecula argue such a remedy may exist in the form of cross-party dialogue about politics, implying the simple act of having a civil discussion may reduce affective polarization and improve the delicate state of the U.S. social fabric.
Existing work on intergroup contact suggests such cross-party dialogue may help reduce partisan animosity through three mechanisms: increased perceptions of commonality, higher rates of perspective-taking, and greater feelings of respect. At the time of this book’s publication, however, there were no applications of this particular intervention toward the end of reducing affective polarization, with some studies even suggesting such interventions could backfire.
Levendusky and Stecula, therefore, perform the first intergroup contact intervention through a lab-based experiment. By testing the efficacy of such an intervention in reducing out-party animosity, the authors provide a baseline from which partisan bridging interventions can be evaluated and deployed by future practitioners, and provide a potential salve to a seemingly dire trajectory of partisan relations in the U.S.
Analytical Approach
The main analysis of the book centers around an in-person cross-party dialogue intervention meant to reduce partisan animosity. The authors recruit (through advertisements on Facebook) a pool of 553 subjects taking part in 1 of 28 sessions from November 2018 to July 2019 in the Philadelphia metro area, who are each compensated $20. All subjects completed a pretest questionnaire of partisanship and demographics and were subsequently assigned to one of three treatment groups (see Table 1).
Table 1: Treatment conditions of Levendusky and Stecula (2021)
| Condition | Group Composition | Discussion Article |
|---|---|---|
| Partisan homogeneous discussion | One party | "Amazingly, our partisan divide is getting even wider" |
| Partisan heterogenous discussion | Both parties | "Not So Divided After All" |
| Apolitical control | Both parties | "Jersey Shore Ranked Best Place for Summer Vacation" |
After the subject had an opportunity to read their article, a general discussion (of roughly 15 minutes) proceeded in groups of four to six, with partisan composition depending on treatment conditions. Importantly, subjects in the partisan homogeneous condition discussed an article about growing partisan divides, while the partisan heterogenous condition (where both parties were represented in the group) discussed an article about partisans agreeing on many policies. Subjects then concluded with a posttest questionnaire, with feeling thermometers for both parties, thermometers for candidates of parties, and evaluations of out-party traits, trust, and measures of social distance. Respondents were contacted for a follow-up survey one week later, which 35 percent completed.
In addition to the main experimental analysis comparing the effects of different discussion groups and articles, the authors also conduct a simple correlational analysis using ANES data from the 2008-2009 panel to determine if any relationship exists between frequency of cross-party discussion and positive out-party sentiment.
Main Findings
Compared to the control group, subjects who engaged in political discussions with out-partisans reported overall less affective polarization, warmer out-party feelings, better evaluations of out-party traits, higher trust in out-partisans, and less desire for social distance. These findings complement results from the correlational analysis, which found respondents with greater self-reported contact with out-party members held warmer evaluations of the opposite party. Those in the homogenous treatment condition, however, showed no effects.
Results are moderated, however, by strength of partisanship. The heterogenous treatment condition had no effect on the attitudes of strong partisans (save evaluations of out-party candidates). Consumption of partisan media, however, had no moderating effect (although the authors are underpowered to detect such an effect with their sample).
The authors also attempt to determine the mechanisms behind such decreases in animosity, measuring post-treatment perceptions of common ground, understanding the other party’s perspective, belief in the other party having reasonable beliefs, and belief in the other party respecting the subject’s beliefs. Across all mechanism checks, those in the heterogenous condition were more favorable of members of the opposite party compared to those in the treatment condition.
Finally, testing the durability of results, the authors found those in the heterogenous treatment resurveyed on week later were still less affectively polarized, rated the out-party more warmly, and were less inclined to engage in social distancing than those in the control condition (although effects waned for evaluations of out-party candidates, traits, and trust).
Implications
The results of this book imply growing partisan animosity need not be an inescapable trend. While certain populations are resistant to appeals for civility (such as strong partisans), it seems some are amenable to both having civil conversations with non-copartisans and updating their previously held beliefs about those non-copartisans in a positive manner.
This book also adds to a growing literature bounding the scope of affective polarization in American politics. While a common perception is that no partisan is willing to engage with the other side, Levendusky and Stecula show this is not the case. A large portion of the American population already does interact with out-party members, and those who were assigned to the heterogenous treatment condition indicated a higher willingness to interact with such out-party members again in future conversations. While by no means a majority of the population, this sizable group could offer an especially fruitful population within which interventions aimed to reduce partisan animosity could have greater effects. Indeed, the book reminds of the non-centrality of partisanship in the everyday lives of most Americans.
Finally, the findings from this book imply potential paths forward through other interventions. Encouraging other personal and politically cross-cutting social connections may help reduce partisan animosity, perhaps through formal organizations or social groups.
Questions left unanswered
While the authors make clear their intervention was effective in reducing partisan animosity, what is unclear is the extent to which the component pieces of the treatment were relatively responsible. The treatment was a bundle of both the article topic and the composition of the discussion group, which were not independently randomized. The authors do separately test just the effect of the article topic on a separate sample and do see some effects of varying similarity, but the effect sizes are not directly comparable.
More broadly, the scope of the effectiveness of cross-party discussions is left unanswered, if only because the authors use their experiment to set a baseline for future work. For example, could such interventions work in online environments, where civility is not guaranteed? This raises the question of what pre-conditions are necessary for the effectiveness of the intervention, which the authors debate heavily in the text of the book.
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
Levendusky and Stecula should be commended on their successful fielding of a non-student pool of subjects willing to commit to a lab-based intervention and a follow-up survey. As they detail, this is no simple task. Of course, the results of this persistence in fielding such a sample is a double-edged sword. While the authors were rewarded with an opportunity to test a unique intervention offline, the pool of respondents the intervention is tested on is inherently unrepresentative. This is not to say the pool was not diverse (as the authors indicate it was), but that the population of Americans willing to sit down for a cross-party dialogue may be attitudinally dissimilar from the U.S. population writ-large. The authors point out that many more people are willing to (and do) engage in such dialogues than are generally perceived, but this would seemingly just set a ceiling on potential intervention scaling to the general population. The authors suggest this is a cheap intervention, at around $3.13 per respondent. It is unclear, however, how widely such an intervention would need to be deployed for there to be any sort of behavioral change in the population. With a voting-age population of roughly 250 million (U.S. Census Bureau), a large-scale deployment would become prohibitively expensive very quickly. This is likely not necessary, but more work needs to be done to determine the desired population-level outcomes of efforts to reduce partisan animosity.
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
Levendusky, M., & Stecula, D. (2021). We Need to Talk: How Cross-Party Dialogue Reduces Affective Polarization (Elements in Experimental Political Science). Cambridge: Cambridge University Press. doi:10.1017/9781009042192
Bibtex
1
2
3
4
5
6
7
8
9
10
@book{levendusky_stecula_2021,
place={Cambridge},
series={Elements in Experimental Political Science},
title={We Need to Talk: How Cross-Party Dialogue Reduces Affective Polarization},
DOI={10.1017/9781009042192},
publisher={Cambridge University Press},
author={Levendusky, Matthew S. and Stecula, Dominik A.},
year={2021},
collection={Elements in Experimental Political Science}
}