Rising affective polarization—animus toward supporters of rival political parties—represents a serious concern in the contemporary United States and worldwide (e.g., Druckman and Levendusky Reference Druckman and Levendusky2019). Affective polarization is growing most sharply among politically engaged strong partisans and political activists—one of the most influential groups in the mass public (Ladd Reference Ladd2018).Footnote 1
In this paper, we argue that being trained in and engaging in a form of partisan campaigning can reduce affective polarization among these activists. In particular, we argue that implementing personal persuasion campaigns that involve two-way conversations with out-partisan voters, such as through door-to-door canvassing, can reduce affective polarization among the implementing activists. Despite how widespread interpersonal persuasion efforts are in American politics and the importance of the population that implements them (Enos and Hersh Reference Enos and Hersh2015), we know remarkably little about their effects on the activists who implement them.
Theoretically, the effects that implementing such a campaign would have on activists are ambiguous. On the one hand, theories of intergroup relations might suggest that such activity would only make affective polarization worse. Political persuasion campaigns are fundamentally about winning intergroup competitions. Although contact between groups is often assumed to reduce prejudice, intergroup contact often exacerbates prejudices in competitive contexts (e.g., Lowe Reference Lowe2021). Consistent with this, Michelitch (Reference Michelitch2015) finds that affective polarization rises during elections.
On the other hand, research on perspective-getting offers reasons to be more optimistic. This research finds that hearing narratives about out-group members can durably reduce prejudice, potentially by humanizing and individuating out-group members (Audette, Horowitz, and Michelitch Reference Audette, Horowitz and Michelitch2020; Kalla and Broockman Reference Kalla and ForthcomingForthcoming). Persuasion campaigns should instruct canvassers to approach potentially persuadable voters in a manner seeking to minimize conflict and maximize these voters’ openness to change. And indeed, many political persuasion campaigns instruct canvassers to pursue such an approach, asking questions of voters who do not agree with them and prompting these voters to talk about themselves. But when canvassers ask these voters to share their views and stories, canvasser perspective-getting may be an unintended byproduct: those who implement these campaigns end up hearing personal narratives from a number of voters with opposing views. Perspective-getting could reduce affective polarization in this context through several mechanisms. First, conversations may humanize out-partisans by revealing experiences to which the canvasser can relate and facilitating emotional connections. Second, such exchanges could individuate out-group members by revealing that individuals who disagree have a variety of individual experiences and reasons for doing so. We argue that this is likely to occur.
To test our argument, we report three within-subjects studies from real voter outreach campaigns that encouraged activists to have in-depth conversations with voters they sought to persuade. In these studies, we analyze changes in affective polarization among the political activists who chose to participate in these campaigns. These campaigns spanned efforts to persuade voters both by phone and door, about both a partisan election and an issue, and both during an active election and outside of an electoral context. As we discuss in greater detail, features of these contexts and the diversity of their time, place, and manner lend support for our statistical identifying assumptions. Across all these contexts, we find that participating activists’ levels of affective polarization considerably declined in the days and weeks following their participation. Furthermore, qualitative responses are consistent with these conversations reducing prejudice by creating opportunities for perspective-getting, wherein activists heard out-partisans’ perspectives, humanizing and individuating them.
Contributions to the Literature
Our research contributes to two literatures.
First, a growing body of research has studied methods to reduce affective polarization (e.g., Levendusky and Stecula Reference Levendusky and Stecula2021; Rossiter Reference Rossiter2021). (This is an extensive literature and, given space constraints, we provide further examples in Online Appendix Table A1.) We build on this research in two ways. First, despite intergroup contact’s prominent place in theories of prejudice reduction, very few existing studies examine the effects of interpersonal contact between partisans in field settings (see Appendix Table A1). Moreover, the few studies that do so generally sideline discussions of political differences, instead focusing on discussions of similarities across party lines (for one exception, see Rossiter Reference Rossiter2021). Second, although affective polarization is most sharply increasing—and its effects are arguably the most politically worrisome—among political activists, little previous research has focused explicitly on this population.
A second body of research has found that interpersonal conversations can be effective tools for campaigns to use to change voter attitudes and behaviors (e.g., Gerber and Green Reference Gerber and Green2000; Kalla and Broockman Reference Kalla and Broockman2020). But this large literature has been entirely focused on the effects of these conversations on the voters campaigns target. We contribute to this literature by providing unique data on how these campaigns influence those who implement them.
It would be good news if interpersonal persuasion campaigns reduced affective polarization among implementing political activists, as these campaigns are incentive-compatible for political elites to implement: the very same tactic that builds support for their cause also reduces affective polarization among their base. Moreover, reducing affective polarization could remove a barrier to democratic deliberation (Chen and Rohn Reference Chen and Rohla2018) among activists, making them more likely to be willing to engage across partisan lines in the future.
Data and Methods
Context
We embedded surveys prior to the baseline training and during endline debriefing materials for political activists implementing three separate voter persuasion programs. Studies 1 and 3 were both preregistered, whereas Study 2 followed the same analysis as we preregistered for Study 1.
We first describe the context of these programs. Studies 1 and 2 were embedded within programs attempting to reduce exclusionary attitudes toward undocumented immigrants. Study 1 (
$ n\hskip0.35em =\hskip0.35em 23 $
) involved door-to-door conversations and was conducted in winter 2019–2020 with a median of 23 days between political activists’ last conversation and when they completed the endline survey. In total, 48% of the conversations were with voters who identified as Republican or conservative. Study 2 (
$ n\hskip0.35em =\hskip0.35em 23 $
) involved phone conversations and was conducted in summer 2020 with a median of seven days between the the last conversation and the endline survey. In total, 44% of the conversations were with voters who identified as Republican or conservative.
Study 3 (
$ n\hskip0.35em =\hskip0.35em 104 $
) took place in the context of a persuasion program conducted by a political organization, People’s Action. Political activists from People’s Action had phone conversations with predicted Republican-leaning voters to persuade them to support Biden over Trump for president.Footnote
2
Additional survey and campaign implementation details are available in Appendix Section C.
Propitious for the generalizability of our conclusions, these studies capture the effects of canvassing across a variety of contexts: as detailed in Appendix Section C, spanning multiple states; campaigns to persuade voters both door-to-door (Study 1) and by phone (Studies 2 and 3); about both a partisan election (Study 3) and a politicized issue (Studies 1 and 2); and both during an active election (Study 3) and outside of an electoral context (Studies 1 and 2). The canvassers for all three studies were paid hourly to canvass.
An important commonality between all of these persuasion programs is that the implementing political activists engaged in two-way conversations with voters. In particular, in all three programs, the political activists would begin the conversations by informing voters that they were there to discuss the political issue at hand. Political activists would then ask voters about their opinion and reasons for it. The political activist would then share a personal story about the issue being discussed. The political activist would also ask the voter if they had a similar story they would like to share. Voters shared a story in 74% of the conversations. After this exchange of narratives, the political activist explained why they were canvassing and why they hoped the voter would become more supportive toward their political cause. The political activist then answered any questions the voter may have and responded to any concerns. Finally, the conversation ended with the political activist asking the voter whether and why the conversation changed their mind about the political issue. These conversations typically lasted for 10 minutes. Canvassers received an initial training and ongoing coaching. Canvassers were instructed to complete the baseline survey prior to the initial training. This approach produced in-depth, two-way conversations conductive to perspective-getting on the part of the implementing political activists, but we stress that our conclusions likely would not generalize to outreach programs that did not have these components.
In none of these studies did we measure affective polarization among the voters, as reducing affective polarization was not the goal of any of the programs; they were focused on changing voter attitudes toward immigrants (Studies 1 and 2) and the 2020 presidential election (Study 3). As reported elsewhere (Kalla and Broockman Reference Kalla and ForthcomingForthcoming), the campaigns were all successful at achieving these objectives, although canvassers were not aware of those results when they completed our surveys.
Surveys
Canvassers were encouraged to participate in the baseline surveys by the implementing organizations as part of their onboarding process, before receiving any training. The baseline surveys gathered canvassers’ email addresses. After the programs finished, we sent the canvassers follow-up surveys. Both the baseline and endline surveys contained a standard feeling thermometer question that asked separately about feelings toward Republican voters, Democratic voters, Trump supporters, and Trump opponents.
A number of factors reduce the possibility that demand effects would contaminate our findings. The endline surveys were clearly not from the organization, but from us. The endline surveys were only gathered after the program (and their employment) had ended. The surveys also promised that we would keep responses anonymous and confidential. Furthermore, the respondents were blind to our hypotheses regarding affective polarization when they completed these surveys. The programs themselves were not intended to reduce implementing activists’ level of affective polarization, nor were they ever described as such to canvassers; they were focused on changing voter attitudes. Finally, the surveys included many other questions, provided by the partner organizations, about political activists’ motivations and reflections.
Demographic details on the political activists who completed both the baseline and endline surveys are presented in Tables A3–A5. Of particular note is that, as one would expect, the political activists in each study were highly politically liberal. For example, 74% of political activists in Study 3 identified as very liberal but only 5% identified as moderate and 0% identified as conservative or very conservative. (Nationally, about 25% of Americans identify as liberal or very liberal and 36% identify as conservative or very conservative.Footnote 3 ) Furthermore, across all three studies, 63% stated that they had previously volunteered or worked for a political campaign, around 10 times the 2016 national average (Ansolabehere and Schaffner Reference Ansolabehere and Schaffner2017), and 22% reported having previous experience canvassing with perspective-getting; this prior exposure may bias our treatment effect estimates toward zero.Footnote 4 In Tables A12–A13 we assess the predictors of responding to the endline survey and find no meaningful differences between the canvassers who did and did not respond to this survey.
Because all three programs largely were implemented by liberal activists who were advocating in a liberal direction (for relaxing immigration restrictions in Studies 1 and 2 and for Joe Biden’s election in Study 3), we consider views toward Trump supporters and Republican voters as capturing attitudes toward out-partisans.
Statistical Model
To estimate the effect of participating in this political canvassing on the attitudes of these political activists, we use paired
$ t $
tests, comparing activists’ endline and baseline responses.Footnote
5
The primary identifying assumption is that political activists’ attitudes would not change over time for other reasons—that is, their levels of affective polarization would not have decreased absent their participation in these programs. We believe this assumption is very likely to hold for several reasons. First, as shown in Table A2, the surveys typically took place over a short time span (of only a few weeks), making it unlikely that other events would have affected their views. Second, as affective polarization has generally increased over time, especially among political activists, we would expect any bias to cut against our findings, as we should see affective polarization increasing over time, not decreasing, due to background changes. Third, the only salient event that took place during one of our studies was the 2020 election that took place during Study 3, but previous research again suggests that this should have made affective polarization increase, not decrease (Michelitch Reference Michelitch2015). (The campaigns in Studies 1 and 2 were not timed to coincide with any particular local or national political events.) Thus, our results may understate the causal effects of program participation. Finally, these studies were all conducted at three different periods yet produced substantively similar results. It seems unlikely that affective polarization would have happened to meaningfully decrease at the three different periods that happened to coincide with when these three projects occurred.
It also bears mention that our estimates are specific to the kind of political activist who will agree to take part in such a program. This is arguably the most relevant group—we estimate the treatment effect of participation among the kind of individuals who actually agree to participate in it. However, we note that this population does not appear to be unusually open to out-partisans; as described below, their average feelings toward out-partisan voters at baseline were still quite negative.
Results
Across all three studies, we find that participating in the canvassing program decreases out-partisan animus and affective polarization.
Figure 1 shows the raw baseline (x axes) and endline (y axes) ratings of Trump voters (Figure 1a) and Republican voters (Figure 1b) among canvassers who completed both surveys. The black lines show 45 degree lines, so dots above the line show cases when individuals’ views toward Trump supporters grew more positive. As can be seen, there was a general improvement in canvassers’ views between baseline and endline.Footnote 6
Figure 1. Raw Data: Baseline and Endline Feeling Thermometer Ratings
Note: Jitter added to all plots.
To more formally test for these differences, we conduct paired t tests, as described above. Figure 2 shows these results (exact numerical values are given in Appendix D).
Figure 2. Results from Paired t Tests
Note: Point estimates from paired t tests are surrounded by one standard error (thick lines) and 95% confidence intervals (thin lines).
First, on the feeling thermometer toward Trump voters, there is an average feeling thermometer rating of 17.9 in the baseline surveys (averaging across studies), quite low. Participating in the canvassing program increases this by a statistically significant 4.5 points on average (
$ \mathrm{SE}\hskip0.35em =\hskip0.35em 1.5;p<0.01 $
). Similarly, on feelings toward Republican voters, we find a statistically significant increase in the feeling thermometer of 3.8 points (
$ \mathrm{SE}\hskip0.35em =\hskip0.35em 1.4;p<0.01 $
) from a baseline of 30.6.Footnote
7
Second, there are similar results when we code the dependent variable as affective polarization, the difference in affect toward one’s own side minus one’s affect toward the other. In particular, the bottom two panels of Figure 2 show that affective polarization declined by 6.0 points on Trump supporters minus opponents (
$ \mathrm{SE}\hskip0.35em =\hskip0.35em 2.5;p<0.05 $
) and by 4.0 points on Democrats minus Republicans (
$ \mathrm{SE}\hskip0.35em =\hskip0.35em 1.9;p<0.05 $
).
We see no signs of change in views toward Democrats or Trump opponents, so these effects on affective polarization are driven by changes in views toward Trump supporters and Republicans. We also see no meaningful changes in views toward Republican or Democratic politicians, consistent with prior work showing that people have distinct attitudes toward political elites (Druckman and Levendusky Reference Druckman and Levendusky2019). Importantly, however, these results show that the results are not due to canvassers who enjoyed participating in the programs simply rating all groups more highly; the effects are specific to evaluations of out-partisan voters.
These results appear largely consistent across the three studies despite their differing contexts. Although large standard errors in some cases mean that the coefficients are not statistically significant in every study, the point estimates are quite similar.
The magnitudes of the effects we observed are also substantively large. Boxell, Gentzkow, and Shapiro (Reference Boxell, Gentzkow and Shapiro2020) find that affective polarization has been increasing at a rate of approximately 0.42 points per year, making our findings equivalent to reversing approximately one to one-and-a-half decades of increases in affective polarization.
Qualitative Evidence on Mechanisms
What is causing these reductions in out-partisan animus? Allowing us to probe mechanisms, the endline survey for Study 3 (the partisan presidential program) asked the political activists “What aspects of this project most surprised you?”
The results are summarized in Table A14, where we coded each response into one mutually exclusive category; 85% gave any answer. Not all canvassers who answered this broad question described experiences with out-partisans, but 59% did. Moreover, the most common categories of responses reflect experiences that come with perspective-getting.
First, 26% of responses noted that the experiences humanized out-partisans through emotional connections formed while exchanging narratives. Second, 12% of political activists noted that getting out-partisans’ perspectives led them to see them as nonmonolithic, heterogeneous individuals.
Less common were respondents noting that they learned they shared values with Republican-leaning and Trump voters. Similar to learning that not all out-partisans are monolithic, this information about shared values surprised the political activists. These last two mechanisms are consistent with research that reducing inaccurate second-order beliefs can reduce polarization (e.g., Ahler Reference Ahler2014).
This open-ended question also revealed negative reactions among about 7% of respondents. Among these respondents, the out-partisans they talked with may have reinforced whatever negative stereotypes they held. However, these negative reactions represent a small minority.
Figure A1 shows the average pre–post differences in Study 3, broken down by open-ended response category. The sample sizes are small, but the largest increases in ratings of out-partisans appear to be among the canvassers who made comments describing out-partisans as less monolithic or who formed humanizing emotional connections.
Discussion
Across three studies, we found that when political activists are trained and participate in personal persuasion programs involving two-way conversations with voters who disagree, this can reduce activists’ own affective polarization.
The linchpin of the improvements we observed appears to be perspective-getting. Prior research found that voters who hear perspectives of out-groups from canvassers grow more tolerant toward those out-groups (Kalla and Broockman Reference Kalla and ForthcomingForthcoming). Our research shows that, in the process of having these conversations, the stories voters tell canvassers back also change the canvassers’ own attitudes in turn.
This study has three limitations worthy of reiterating and that future research can examine. First, we do not know whether the reductions in affective polarization we observed were permanent. Across Studies 1 and 2, a median of 15 days passed between activists’ last conversations with voters and their participation in the endline survey (we do not have these data for Study 3). This is a long period relative to the immediate outcome measurement that took place in many of the lab and survey experiments in previous literature (see Table A1), but nevertheless it leaves open whether the reductions are more permanent. Second, this study relied on a within-subject design, although in our settings it is highly unlikely that affective polarization would have gone down for some other reason during all three of the separate periods when the studies were conducted. This research design also limited our ability to definitively probe mechanisms. Third, this study was limited to examining reductions in affective polarization among liberal political activists. Although prior work on other issues finds that perspective-getting can shift issue attitudes among conservative voters (Kalla and Broockman Reference Kalla and ForthcomingForthcoming, Table OA36) and related work from Baron et al. (Reference Baron, Blair, Choi, Gamboa, Gottlieb, Robinson, Rosenzweig, Turnbull and West2021) finds that conversations among college students reduces affective polarization, we do not know whether our findings would generalize to more conservative populations or to the general population. Nevertheless, our results suggest both theoretical and practical lessons.
First, on a theoretical level, our findings further bolster perspective-getting as a paradigm for prejudice reduction that operates through different mechanisms than contact theory. Much research on prejudice reduction follows from Allport’s (Reference Allport1954) insights that common goals and intergroup cooperation help facilitate intergroup comity, whereas competition does the opposite. Prior research on perspective-getting has shown that hearing perspectives from out-group members can reduce out-group prejudices even without having contact with an out-group member (e.g., Kalla and Broockman Reference Kalla and Broockman2020; Reference Kalla and ForthcomingForthcoming). The present research now shows it can do so in a context suffused with intergroup competition. Therefore, these findings support an additional paradigm for thinking about prejudice reduction that places greater emphasis on individuals’ emotional responses and perceptions of individual out-group members.
Second, our results suggest a set of lessons for political campaigns. As noted above, not all political campaigns conduct personal persuasion programs that center on facilitating perspective-getting. Prior research suggests that introducing perspective-getting may allow these campaigns’ conversations to more effectively achieve their goal of persuading voters (Kalla and Broockman Reference Kalla and Broockman2020; Reference Kalla and ForthcomingForthcoming), but our findings suggest they may also make these conversations more beneficial for campaigns’ own staff and volunteers, not to mention potentially for American democracy. As political campaigns already conduct personal outreach on a tremendous scale, introducing opportunities for perspective-getting into more political outreach represents an incentive-compatible approach for reducing affective polarization, helping campaigns more effectively achieve their goals in a manner that also may have benefits for democracy.Footnote 8 How to further scale these efforts or create other scalable opportunities for partisans to engage in perspective-getting is an important question for future research.
Supplementary Materials
To view supplementary material for this article, please visit http://doi.org/10.1017/S0003055422000132.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/CXJBEK.
ACKNOWLEDGMENTS
We thank Chris Bail, Jamie Druckman, Julia Minson, Samara Klar, and Matthew Levendusky for helpful feedback. Programmatic support was provided by Kimberly Serrano of the California Immigrant Policy Center, Adam Kruggel of People’s Action, and Ella Barrett and Steve Deline of the New Conversation Initiative. All remaining errors are our own.
FUNDING STATEMENT
This research was funded by the Immigration Strategic Messaging Project and People’s Action.
CONFLICT OF INTEREST
The authors declare no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The authors affirm that this article adheres to the APSA’s Principles and Guidance on Human Subject Research. The authors declare the human subjects research in this article was reviewed and deemed exempt by the Yale University Human Subjects Committee.
Open access
Comments
No Comments have been published for this article.