Priming Norms to Combat Affective Polarization

Abstract

The American public has affectively polarized such that partisans increasingly dislike the “other side,” and this may have deleterious consequences for a representative democracy. Yet, efforts to reduce partisan hostility arrive at mixed results. We propose a new approach that involves strategically priming civic norms with language tailored to a target audience. We argue that emphasizing group-based civic norms that invoke an “obligation to others” can reduce out-party animus. We test this approach on an important subgroup: U.S. military service members. Like the broader American public, service members have unfavorable feelings toward the opposing party, and these feelings appear to have become more negative in recent years. We use a survey experiment to demonstrate that priming an obligation to others civic norm attenuates affective polarization. Our study advances public opinion research on an understudied subgroup of the population, but more importantly, the theoretical argument has implications for addressing polarization and partisan discord among the mass public and other subgroups.

Keywords

affective polarization civic norms partisanship military priming

While scholars debate the extent to which Americans are divided in their policy opinions (Abramowitz 2010; Fiorina et al. 2008), there is considerable evidence that the mass public has “affectively” polarized such that partisans increasingly dislike the “other side” (Abramowitz and Webster 2016). A substantial literature analyzes the origins of affective polarization and its adverse consequences (e.g., Iyengar et al. 2012; Iyengar et al. 2019; Rogowski and Sutherland 2016). Affective partisan divisions lead to prejudice against opposing partisans (Lelkes and Westwood 2017) and can color apolitical decisions, such as economic choices (McConnell et al. 2018) and evaluations of university scholarship applications (Iyengar and Westwood 2015). Further, there are broader repercussions of partisan discord for governance in a representative democracy. Levendusky (2017) suggests that affective polarization “makes governance more difficult” (59), and Iyengar and Krupenkin (2018) assert that the implications for “political accountability are cause for concern” (214).

Research on tactics to depolarize the mass public has mixed results. Some strategies fail to reduce affective partisan divisions (e.g. Levendusky 2018), and other efforts, though more successful, may not resonate with particular subgroups. For example, Levendusky (2017) uses national samples to demonstrate that priming a common ingroup identity—“Americans’ national identity”—attenuates affective polarization, but this approach may not constrain partisan animus among individuals who already have a heightened sense of American identity.

We offer and test a distinct theoretical approach rooted in group-based civic norms that emphasize an “obligation to others” to address affective polarization. We argue that priming civic norms, “the shared set of expectations about the citizen’s role in politics” (Dalton 2008, 78), can mitigate affective polarization, and this is particularly effective when norms are tailored to a target audience and invoke an obligation to others in the context of politics and partisanship. Emphasizing civic norms reduces the effects of partisan endorsements and increases willingness to discuss politics with opposing partisans (Mullinix 2018). Heightening civic norms stimulates open-minded information search (Kam 2007) and impacts updating of partisan identities (Groenendyk 2013). Moreover, psychology regularly highlights the power of social norms to reduce prejudice toward outgroups (e.g., Monteith et al. 1996). Notably, none of these studies test whether priming norms reduces affective partisan divisions.

To test the potential of group-based civic norms to reduce affective polarization, we selected an important subset of the public: U.S. military service members. For simplicity, we refer to both former service members (veterans) and current service members (active-duty) as “service members” unless otherwise noted. We chose this subgroup for a few reasons. First, the military is an ideal community to test our argument that priming civic norms reduces partisan hostility. Like the broader American public, service members have affectively polarized in recent years—a finding we show using ANES data. At the same time, however, there is a longstanding “apolitical” civic norm within the military which holds that the military has an obligation to all Americans to avoid public partisanship (Garamone 2016; Shanahan 2019); it stands “above the dirty business of politics” (Kohn 2002, 27). This group norm is consistent with other civic norms in that it establishes normative expectations of political behavior and an obligation to others. Consequently, we see an opportunity to prime this civic norm using group-specific messaging directed at an affectively polarized population. Second, by including former service members in our study, we can explore whether the norm resonates with people after they have left military service. Although the norm may be more salient for current service members than veterans, research suggests social norms are resistant to change (e.g., Young 2015). Finding that the norm can influence those who are no longer serving would provide strong evidence that social norms are both influential and durable. Finally, reducing affective polarization has practical importance for the military. Partisan animosity could undermine morale, teamwork, and relationships with civilian policymakers. Successful depolarization strategies may lessen these effects.

After establishing our theoretical argument, we present the results of a survey experiment implemented with over 1000 military service members. We find that priming a group-based civic norm rooted in an obligation to others increases their positive feelings toward the out-party and their willingness to associate positive traits with opposing partisans. The norm also moderates the influence of partisanship on how difficult service members believe it is to serve alongside and under the command of opposing partisans. Our results also suggest there are limitations in appealing to common ingroup identities as a strategy to ameliorate partisan divisions. We find no evidence that priming national identity reduces service members’ partisan divisions; this underscores the importance of identifying new theoretical approaches to address partisan animosity. Although this norm and message are specific to the military, civic norms that invoke an obligation to others are not. This study is a “proof of concept” test of our theoretical argument about the potential of group-based civic norms to combat outgroup hostility. In our conclusion, we discuss how our theoretical argument may generalize and the types of civic and social norms that may reduce affective polarization among other groups and the mass public.

Affective Polarization

Affective polarization is the “tendency of people identifying as Republicans or Democrats to view opposing partisans negatively and copartisans positively” (Iyengar and Westwood 2015, 691). It is a result of people’s identification with a political party. Using the theoretical lens of social identity (e.g., Tajfel and Turner 1979), scholars suggest that people divide the world into partisan ingroups and outgroups which, in turn, prompts positive feelings toward their ingroup, and negative evaluations of the outgroup (see Iyengar et al. 2019). A substantial literature discusses the causes of rising affective polarization in the United States. (Iyengar et al. 2012; Iyengar and Krupenkin 2018; Rogowski and Sutherland 2016), and it is clear that the extent of partisan hostility has fundamentally changed.

Affective polarization is conceptualized in different ways. For many researchers, feelings are at the root of the concept. To capture feelings toward each party (and the gap between the two), scholars employ feeling thermometers where people rate how warm or cold they feel toward the parties. Affective divisions are also revealed in the stereotypes that partisans have of each other. As such, some researchers use trait ratings to determine the extent to which people view the opposing party as intelligent and honest or hypocritical and selfish. A more contentious conceptualization is the use of social distance measures which gauge how comfortable people are with having opposing partisans as friends, neighbors, or in-laws (see Klar et al. 2018 for a critique).¹ Across measures, scholars not only document the rise of affective polarization, but they also demonstrate its consequences for ingroup favoritism and outgroup discrimination (e.g., Iyengar and Westwood 2015; McConnell et al. 2018).

Sparked by concerns about the consequences of affective polarization, research explores—with mixed success—methods to mitigate the partisan discord (Ahler and Sood 2018; Levendusky 2017, 2018; Wojcieszak et al. 2020). Levendusky (2018), for example, finds that self-affirmation techniques and strategies that heighten people’s partisan ambivalence are ineffective and can even “backfire” and increase the partisan divide. Levendusky (2017) has more success using a common ingroup identity to address intergroup bias. He demonstrates that priming a “national identity”—getting people to think of themselves as “Americans” rather than just “Democrats” or “Republicans”—lessens affective polarization. Yet, this strategy may not be effective for everyone. In a manipulation check, Levendusky finds that the prime only increases respondents’ American identity from 4.3 to 4.4 on a 5-point scale. He notes that this modest shift, although statistically significant, is “due to a ceiling effect” (62). This suggests that this prime may not be effective for subgroups that have a higher baseline sense of national identity.

We propose an alternative strategy rooted in a different theoretical argument. Rather than trying to shift people’s sense of national identity, we see potential in emphasizing group-based civic norms to constrain partisan animosity.

Priming Norms to Combat Affective Polarization

We argue that priming specific civic norms can reduce affective polarization. Civic norms are shared expectations for how a person should behave in democratic politics. They are “what people think people should do as good citizens” (Dalton 2008, 78). More broadly, social norms (which may or may not be civic in nature) can be descriptive and signal what is normal behavior, and they can be injunctive and refer to “rules or beliefs as to what constitutes morally approved and disapproved conduct” (Cialdini et al. 1990, 1015). Norms tell us both what “is” typical behavior as well as what “ought” to be done. Scholars unpack different types of civic norms (Conover et al. 1991; Dalton 2008), and the specific norms and expectations vary across individuals, groups, and contexts.

While the political science literature focuses on their effects for political participation (Blais 2000; Campbell 2006), civic norms also influence political preferences. When they are made salient, civic norms alter relevant attitudes and behaviors. Kam (2007) finds that reminders of the norms and duties of citizens lead people to learn more about candidates’ issue positions and engage in more open-minded information search. Groenendyk (2013) demonstrates that people update their partisan identities in response to new information when civic norms are made salient. Other work shows that priming civic norms leads people to shirk party endorsements in argument evaluations and makes them more willing to discuss policies with people from the opposing party (Mullinix 2018). Both Kam and Mullinix argue that civic norms map onto a notion of “obligation to others” which is linked to accountability and cognitive effort in decision-making (e.g., Tetlock 1983). These findings suggest that, (1) when primed, civic norms transform information processing and the influence of partisanship on decision-making and (2) norms may be particularly impactful when they emphasize an obligation to others. However, none of these studies look at the effects of priming norms for affective polarization.

While political science has only begun to tap into the potential of norms to alter outgroup political attitudes, psychology has firmly established that social norms influence attitudes toward outgroups. For example, Crandall et al. (2002) build on group norm theory to demonstrate how social norms suppress prejudice, where prejudice is a “negative evaluation of a group or of an individual on the basis of group membership” (359). And, manipulation of social norms, their salience, and the degree of consensus surrounding a norm impacts attitudes and prejudice toward outgroups (Monteith et al. 1996; Paluck 2009; Stangor et al. 2001). As such, activating particular civic norms could plausibly reduce prejudice and hostility toward partisan outgroups.

Although not focused on an obligation to others, Wojcieszak et al. (2020) advance a similar argument about norms and affective polarization. They find that manipulating a social norm of “open-mindedness” alters people’s selection of media. The authors test whether this norm mediates indicators of affective polarization. The results are mixed. Although the norm of open-mindedness increases exposure to balanced information in one study, this does not lower affective polarization. In a second study, they find a mediating effect whereby activating the open-mindedness norm impacts ingroup favorability ratings through increased exposure to balance and counter-attitudinal news. Wojcieszak et al. emphasize that “future work needs to more systematically test the effects of various social norms” (22). We agree. Yet, to uncover this it might be necessary to tailor messaging to the target audience and focus on norms with particular characteristics. Rather than focus on open-mindedness, we build on civic norms rooted in an obligation to others.

While we could apply this argument to several target audiences, we focus on U.S. military service members for three reasons. First, the U.S. military maintains a group-based civic norm that encourages service members to avoid “partisanship of word or deed, activity, or affiliation” (Kohn 2002, 27) and emphasizes an obligation to the American public. The military may be a “most-likely” case to test hypotheses derived from our theory (Rohlfing 2012). Second, we are interested in whether civic norms are as persistent as other social norms (Young 2015), and former service members are an ideal population to explore this idea. Finally, while discovering successful depolarization strategies has broad implications for the American public, it is especially important for the military where partisan hostility could undermine the unit cohesion and teamwork that are critical to organizational success.

Before turning to the specifics of this civic norm and our argument, it is important to establish the extent of affective polarization in the military.² Previous research finds that affective polarization is primarily evident in out-party animus rather than in-party affinity (Iyengar et al. 2019). Using ANES data, we compared the mean out-party feeling thermometers of those who served in the military and those who have not.³ The results show that like the broader American public, service members’ feelings toward the opposing party have become more negative in recent years (Table 1).

Table 1.

Mean Out-Party Feeling Thermometers.

	No Military Service	Military Service
2004	39.51	39.82
2008	34.89	33.59
2012	28.11	26.22
2016	26.22	24.99

Note. Results based on weighted ANES data. Thermometer ratings range from 0 to 100. Ratings between 0 and 50 mean that they don’t feel favorable toward the group and that they don’t care too much for that group.

Although the military community resembles the broader public in terms of affective polarization, we believe that the key to reducing this hostility lies in priming specific civic norms that will resonate with this subgroup. More specifically, we think that priming the military’s “apolitical” civic norm can reduce affective partisan divisions.

The military’s apolitical norm is a civic expectation about how service members ought to behave.⁴ Although the Department of Defense (DoD) restricts some political activities (US Department of Defense 2008), the apolitical norm goes beyond those regulations. The norm refers to a military tradition of keeping partisan politics private. Military leaders encourage service members to exercise their responsibilities as citizens, including voting, but in ways that do not imply partisan endorsement by DoD (Shanahan 2019). The norm emphasizes the military’s obligation to defend the Constitution, obey the orders of elected officials, and serve and defend all Americans, not just those from one party. Discussing this norm among service members, former Chairman of the Joint Chiefs of Staff, Marine Corps General Joseph Dunford stated, “The American people…have to look at us as an apolitical organization that swears an oath to the Constitution of the United States—not an individual, not a party, not a branch of government” (Garamone 2016). That the norm is often linked to an “obligation to others” is important given that research on civic norms highlights the impact of this for openness to information and counter-arguments (Mullinix 2018).

The apolitical norm is a key part of the military’s professional character (Betros 2001; Holsti 2001). Huntington (1957) discusses the tradition and argues that a professional, apolitical military is essential to healthy civil-military relations in a democratic society. Although some scholars argue the apolitical norm is eroding (Griffiths and Simon 2019), the apolitical tradition nevertheless remains a central part of the military’s professional ethos. Indeed, about a week after we implemented the survey for this study, Secretary of Defense Mark Esper urged DoD personnel to “uphold DoD’s longstanding tradition of remaining apolitical” (Esper 2020).

We believe that priming service members to think about the apolitical norm will attenuate affective polarization by reducing outgroup prejudice and hostility. Priming a civic norm with language that appeals to service members should increase the salience of the belief that they have an “obligation to others” to stay out of partisan politics. Their oath includes an obligation to support and defend the U.S. Constitution, and the broader norm includes apolitical expectations about defending the American public and maintaining their confidence. Although there may be a latent belief in the norm, priming should remind service members of these obligations. This should, in turn, elevate their positive feelings toward the opposing party.

Importantly, we do not expect the norm to change feeling thermometer ratings of the in-party. Iyengar et al. (2012) argue that affective polarization is largely a result of increasing out-party hostility rather than in-party affinity. Since the norm allows for private political expression, priming the norm should not reduce in-party affect. Additionally, other successful strategies increase out-party favorability rather than reducing in-party affect (Ahler and Sood 2018; Levendusky 2017). Levendusky’s national identity treatment elevates thermometer ratings of the out-party, but it has no effect on people’s feelings toward their own party.⁵ We expect similar results. Heightening the salience of an obligation to others should increase positive feelings toward the out-party without damaging people’s feelings toward their own party.

Additionally, we expect that priming the apolitical norm will increase the positive trait ratings of opposing partisans, but we expect a weaker impact on negative trait ratings. The reason is negativity bias. It is easier to persuade people to believe the other partisan side has positive traits than it is to convince them the other side is not hypocritical, selfish, and mean. Iyengar and Krupenkin (2018) discuss the voluminous literature on negativity bias (e.g., Fiske 1980; Klein 1991; Ohira et al. 1998). They summarily state that “person perception is more sensitive to negative rather than positive traits” and that “negative information is more likely to draw and hold attention, thereby becoming more memorable than positive information” (212). Indeed, Levendusky’s (2017) findings appear to reflect negativity bias. His national identity treatment increases positive trait ratings of opposing partisans but has no effect on negative trait ratings. He reflects, “I lack a clear theoretical expectation as to why this is the case, but it presents an interesting possibility of differential effects for future work” (63). We do not believe his pattern of results is produced by chance, and we expect a similar response to our apolitical norm prime. We suspect that beliefs about negative traits of out-partisans are more strongly held than beliefs about positive traits and thus less likely to change in response to new information (e.g., Taber and Lodge 2006).

Hypothesis

Priming an obligation to others with this group-based civic norm should reduce affective partisan divisions among military service members. Group norms describe common behavior and communicate the expectations for members, and as such, exert pressure on attitudes. Invoking social norms suppresses prejudice and outgroup animus (Crandall et al. 2002). We apply this insight to civic norms and affective partisan polarization. Reminding service members of an apolitical tradition and an obligation to defend all Americans and support the Constitution should mitigate animosity toward opposing partisans. The salience of any norm varies between individuals and by context, but “norms should motivate behavior primarily when they are activated (i.e., made salient or otherwise focused on)” (Cialdini et al. 1990, 1015). Furthermore, experiments reveal that other civic norms can be activated through text-based stimuli (Groenendyk 2013; Kam 2007). Taken together, it seems possible to raise the salience of an apolitical norm among military service members through a text-based prime and that doing so will reduce affective partisan divisions. As noted above, these effects should be most evident in feelings toward and positive trait ratings of the opposing party. We hypothesize that:

Priming an apolitical norm will reduce affective polarization among military service members. These effects will be evident for out-party feeling thermometer ratings and associating positive traits with the opposing party.

Although not our primary focus, we also explore whether priming national identity affects service members. Priming national identity reduces affective polarization among the public-at-large (Levendusky 2017), but if this identity is already salient for a subpopulation, ceiling effects likely constrain efforts to prime it. Military service members may be such a group. We analyzed a national identity measure from the 2016 ANES which asked, “How important is being American to your identity?” We find that, even after controlling for other relevant factors, military service is a statistically significant predictor of the importance of national identity (p < 0.05) (Appendix page 25). Given that service members have an elevated sense of American national identity, it is unlikely that a prime would heighten it. To examine this supposition, we explore the effect of a national identity prime on service members. This allows a comparison with our main treatment (i.e., apolitical norm) and connects our findings to previous literature. If priming national identity has muted effects, it bolsters the importance of identifying other strategies—such as emphasizing civic norms—to address affective polarization.

Research Design

To test our hypothesis, we contracted with Qualtrics to draw a sample of current and former military service members to participate in an online survey (N = 1075). Participants indicated to Qualtrics that they either currently serve or had previously served in the U.S. military. Because our target population was service members, we wanted the sample to reflect characteristics of that population. This is challenging because the distribution of demographic characteristics varies by military status (veteran vs. active-duty) and across branches of the armed forces (see U.S. Department of Defense 2018), and we wanted to include all branches and about an equal number of veterans and active-duty service members. Qualtrics used quotas that resulted in a sample that, like the military itself, skewed toward Republicans and males. The sample yielded 635 Republicans and 438 Democrats.⁶

Respondents were randomly assigned to one of three conditions (see Appendix pages 3–6 for stimuli). To situate our study in prior scholarship, we used the same materials as Levendusky (2017) for our first two conditions. The control group received a news article on a topic unrelated to politics and was prompted to write about a non-political topic. The second and third groups received experimental treatments. Both of our experimental manipulations employ a two-part prime: a text-based stimulus and a prompt to write about the topic. The second group (national identity) received an article that discussed reasons why people love America. They were then asked to write what they thought was the most important reason people like America and why they were proud to be an American.

The third group (apolitical norm) constitutes our main manipulation and was designed to prime this group-based civic norm. People in this condition received an article of the same length as the national identity condition that emphasized a longstanding norm of political neutrality in the military. The message noted that “military service members swear an oath to support and defend the Constitution of the United States” and that “they agree to obey the lawful orders of civilian authorities regardless of which political party holds power.” It quoted prominent military leaders and noted that the military is “obligated to defend all Americans regardless of their political beliefs.” To activate the norm, we invoked words like “norm,” “tradition,” “duty,” and “obligation”; the latter was mentioned three times. We designed the norm to be both descriptive of common behavior and injunctive about how people “ought” to behave. This group was asked to write what they thought was the most important reason the military should be apolitical. This apolitical norm message is consistent with prior descriptions of civic norms in that it communicates “what people think people should do as good citizens” (Dalton 2008), but it is specific to the civic expectations of service members.

Two steps were taken to ensure that our treatment message activated the intended norm. First, we pretested our stimuli with 95 service members (Appendix page 2; N varies by question). A majority indicated that the message would effectively increase awareness of the issue of political neutrality. Indeed, only 7.37% thought the message would be “not effective at all.” Additionally, more than 70% of respondents said that including quotes from prominent military leaders would “probably” or “definitely” increase the persuasiveness of the message. Second, in our main study we included four questions to capture the salience and beliefs about the apolitical norm (post-treatment and after dependent variables). Three of the four measures yield a significant increase in beliefs about the importance of the norm relative to the control condition (p < .05) (see Appendix pages 9–10).⁷ As an example, one question tests whether we activated an obligation to others in the context of politics; we asked “To what extent do you agree or disagree that U.S. military service members have an obligation to the American people to stay out of politics?”

As noted above, it seems plausible that ceiling effects constrain the national identity prime in a military sample. We included three national identity measures (Huddy and Khatib 2007) that were asked after stimuli and dependent variables. Among the control group, two measures yield means of 4.5 on the 5-point scale, and the third produces an average score of 4.1. The identity is already salient; there is limited space to make it more so. For a direct test, we perform ordered logistic regressions predicting each national identity measure by experimental condition (Appendix pages 11–12). In regressions without controls, the prime has a null effect on national identity (Identity Q1 = p < .54; Q2 = p < .83, Q3 = p < .22). An OLS regression using an index based on the average of the three items also has a null effect (Index = p < .31). Including control variables does not change these inferences. It is possible that an alternative national identity measure with more precision at the upper-end of the scales could detect treatment effects that we do not observe here. If, however, the national identity prime fails to shift the salience of this identity as our evidence suggests, it is unlikely to influence affective polarization.⁸

The survey was fielded January 17–29, 2020. The time period is noteworthy because it was after the U.S. House of Representatives voted to impeach President Donald Trump, but the Senate had not yet voted to acquit. We suspect that this time frame is characterized by amplified partisan discord. Thus, we hesitate to draw descriptive inferences about the precise extent of affective polarization among service members. However, we have likely established a conservative test for inferences about average treatment effects; any heightened partisan animosity likely made it more difficult to reduce affective polarization.⁹

We measure affective polarization using two established techniques (e.g., Druckman and Levendusky 2019). First, we employ feeling thermometers (0–100) toward both the Democratic and Republican parties (separately). Second, we use trait ratings. Respondents indicated how well eight traits described out-partisans on a five-point scale that ranged from “not well at all” to “extremely well.” We built positive and negative trait indices and measured the difference between indices. Together, these measures allow us to test if invoking the norm increases positive feelings toward the out-party and prompts people to more closely associate positive traits with opposing partisans. Feelings toward the in-party and negative trait ratings of out-partisans should show less movement. We expect a reduced gap in feelings toward each party, and less difference between the average positive and negative trait ratings of opposing partisans.

As an additional test, we embedded a two-condition party cue experiment within the larger experiment to assess the extent to which the apolitical norm moderates partisan divisions in a military-specific context. Participants were randomly assigned to one of two sets of questions. One group was asked, “How difficult would it be for you to work closely in a military unit with someone who is a strong Democrat?” and “How difficult would it be for you to serve under a military commander who is a strong Democrat?” The second group was asked the same set of questions, but the target was a “strong Republican.” The five response options ranged from “not at all difficult” to “extremely difficult.” The random assignment used for the party cue conditions is distinct from the feeling thermometers (asked of both parties) and the trait rating (asked only about out-partisans). This experiment allows for two analyses. First, it provides another test of affective polarization; in some respects, it is akin to social distance measures of hostility toward partisan outgroups used in prior literature (e.g., Druckman and Levendusky 2019). The only difference between the two conditions is the partisanship of the question target. If we observe a difference in the average reported difficulty, we know that the partisanship of military commander and/or fellow servicemember is what triggered the difference. Second, by embedding this experiment within the larger study, we can see whether the national identity and apolitical norm treatments moderate the party cue effect. We expect affectively polarized service members to report greater difficulty serving alongside and under the command of opposing partisans; however, priming the norm should moderate this effect.

For each measure (feeling thermometers, trait ratings, and party cue experiment), we examine evidence of affective polarization and test the effects of the norm and identity treatments for partisan divisions.¹⁰ Again, our primary focus is on the effect of the apolitical norm. To test our hypothesis, we primarily employ OLS regressions. Following Levendusky (2017), we used a two-part measurement of partisanship. Respondents were first asked, “Generally speaking, do you think of yourself as a Republican, Democrat, or Independent?” People who selected Independent were prompted with a follow-up about if they felt closer to a particular party. Partisan-leaners are treated as partisans. We also perform randomization inference analyses as well as regressions with controls for military service status (veteran or active-duty) and rank, gender, race, age, income, education level, and political knowledge (Appendix pages 13–17).

Results and Analysis

Feeling Thermometers

The results of the national identity and norm treatments for feelings toward both parties are shown in Table 2. The control group is the baseline for comparison. The constant coefficients show that this military sample is affectively polarized. The gap between the baseline in-party and out-party thermometers is almost 48 points on these 0 to 100 scales. As hypothesized, however, the apolitical norm cuts into the divide. Priming the norm significantly increases positive feelings toward the out-party (p < .02), and as expected, it has a null effect on the in-party feeling thermometer (p < .20). The net result is that it reduces the distance between the in and out-party feeling thermometer ratings by over seven points (p < .02). By contrast, the national identity treatment has no statistically significant effect. We are cautious to draw firm conclusions about these null results because failure to detect treatment effects may be driven by many factors. Still, taken together, these results support our hypothesis that the apolitical norm reduces affective polarization.

Table 2.

Feeling Thermometer Results.

	(1)	(2)	(3)
Experimental Condition	Out FT	In FT	FT Distance
National identity treatment	1.802 (2.136)	0.204 (1.581)	−1.598 (2.773)
Norm treatment	5.237** (2.225)	−2.149 (1.653)	−7.386** (2.917)
Constant	31.26*** (1.491)	79.25*** (1.087)	47.98*** (1.928)
Observations	1073	1073	1073
R-squared	0.005	0.002	0.007

Two-tailed tests of significance: *** p<.01, ** p<.05, * p<.1

Note. Coefficients are based on OLS regression models. Positive coefficients reflect “warmer” feelings. The baseline is the control condition. Robust standard errors in parentheses.

Trait Ratings

Next, we evaluate the influence of the treatments on trait ratings of opposing partisans. We asked how well the following traits described Democrats/Republicans: American, intelligent, honest, open-minded, generous, hypocritical, selfish, and mean (order randomized). These questions were only asked about each respondent’s out-party. We isolate the American trait given how it might, intuitively, be shaped by the national identity treatment. We also include it in the positive traits index, but omitting it does not change inferences.

The results of the national identity and norm treatments on opposing partisan trait ratings are shown in Table 3. The constant coefficients again show affective polarization. The difference between the positive and negative trait rating indices is 0.87 on the 5-point scale.

Table 3.

Traits Associated with Out-Party.

	(1)	(2)	(3)	(4)
Experimental Condition	American	Positive Traits	Negative Traits	Trait Index Difference (Neg-Pos)
National identity treatment	0.0345 (0.0982)	0.0491 (0.0860)	−0.00741 (0.0778)	−0.0565 (0.129)
Norm treatment	0.202** (0.102)	0.238*** (0.0891)	−0.0675 (0.0795)	−0.306** (0.130)
Constant	3.131*** (0.0689)	2.704*** (0.0591)	3.578*** (0.0545)	0.874*** (0.0882)
Observations	1073	1073	1073	1073
R-squared	0.004	0.007	0.001	0.006

Note. Coefficients are based on OLS regression models. Positive coefficients reflect a higher level of how well the trait applies to opposing partisans. The baseline is the control condition. Robust standard errors in parentheses.

Two-tailed tests of significance: *** p<.01, ** p<.05, * p<.1

Turning to the manipulations, we hypothesized that because of negativity bias, beliefs about the negative traits of out-partisans are more strongly held than positive traits. Consequently, the positive trait evaluation may be more susceptible to manipulation. This is exactly what we find with the norm treatment. Priming the norm significantly increases how well partisans think the word “American” (p < .05) and other positive traits describe members of the opposing party (p < .01). The coefficient is negative for the effect of the norm for negative trait ratings, but it is not significant (p < .40). Model 4 in Table 3 reveals that the norm has the net effect of significantly narrowing the gap between the average rating for the positive and negative trait indices (p < .02). This bolsters our argument that depolarization strategies affect trait ratings asymmetrically due to negativity bias, but future research should explore the generalizability of the insight. In comparison, the national identity treatment has no statistically significant effect on either the positive or negative trait indices. The American trait rating is akin to a manipulation check of the national identity prime, but as expected, the prime fails to shift this attitude.

Party Cue Experiment

We embedded a two-condition party cue experiment after the dependent variables analyzed above for a few reasons. First, it allows us to test affective polarization using a military-specific measure. Second, it provides a unique test as to whether the norm mitigates affective polarization as hypothesized. Finally, the feeling thermometers were asked of both parties, and the trait ratings were asked only of out-partisans. We saw value in the leverage afforded by random assignment (i.e., embedding an experiment within the larger experiment). Respondents were randomly assigned to one of two sets of questions (strong Democrat vs. strong Republican target). They were asked about how difficult it would be to work closely in a military unit with someone who is a strong Democrat/Republican and how difficult it would be to serve under a military commander who is a strong Democrat/Republican. We categorize respondents by whether they received an in-party or out-party cue question for the analysis and perform regressions interacting the in/out-party cue manipulation with the broader experimental conditions (control, identity, or norm). The results are shown in Figures 1 and 2.

Figure 1.

Difficulty of working with out-party relative to in-party.

Figure 2.

Difficulty of serving under out-party relative to in-party. Note: The figures show the change in predicted difficulty of working with or serving under a commander from the opposing party relative to the in-party by experimental condition. The horizontal dashed line is the in-party cue condition. Positive values indicate an increase in reported difficulty of working with the individual. The vertical lines represent 95% confidence intervals. The figures are based on regression models in the Appendix (page 16).

The effect on the left side of each figure is the predicted effect on expressed “difficulty” when people receive the out-party target relative to the in-party target among respondents in the “control” group in the broader experiment (no national identity or norm prime). That is, the left side of each figure shows the effect of the party cue uncontaminated by our main treatments. Here, the party cue increases the reported difficulty of working with people from the out-party from 1.75 to 2.4 on a 5-point scale (p < .01), and serving under the out-party increases from 1.67 to 2.49 (p < .01). This suggests that affective polarization spills over into the workplace for service members whereby they think it is more difficult to serve alongside or under the command of a servicemember from the opposing party. The figures are based on the full sample, but the inferences do not change if we restrict the analysis to active-duty respondents. While this disconcerting finding is consistent with evidence that affective polarization impacts non-political judgments (Iyengar and Westwood 2015), there are grounds for optimism in that the average out-party rating does not exceed the midpoint (3) which was labeled as “somewhat difficult.” Additionally, we note that the party cue is the only information provided about the servicemember in the question, that is, people are only provided a single attribute (see Dafoe et al. 2018). It is possible that additional information (experience, demographics, etc.) could reinforce or undercut the influence of partisan affiliation.

While the partisan cue effect is interesting, our main question is whether the treatments moderate it. The national identity prime has no moderating effect on the party cue treatment. However, the interaction between party cue condition and the norm is significant for both outcomes (Fig. 1 p < .04; Fig. 2 p < .01).¹¹ The right panel of both figures shows that the norm moderates the party cue effect in both evaluations. Figure 1 reveals that for people who receive the norm, the party cue effect is not significant for reported difficulty of working closely in a military unit with someone from the opposing party (p < .07). And, while our norm cuts into affective polarization, Figure 2 shows that it does not eliminate the effect of the target’s party with respect to perceived difficulty of serving under the individual (p < .03). Thus, priming the norm moderates affective polarization, but it is not a panacea.

Additional Analyses

Thus far, we have focused on the effects of the experimental manipulations. If our argument is correct—that a heightened apolitical norm diminishes partisan discord—this should also be evident in observational analyses. That is, individuals with higher levels of support for the norm in their day-to-day lives should hold warmer feelings toward the out-party and more strongly associate positive traits with opposing partisans.

To test this possibility, we restrict our analysis to the control group; individuals for whom we did not manipulate the norm or national identity. We built an index of the norm by standardizing the three manipulation check questions that were successfully manipulated by our norm treatment. We then regressed our two main outcome variables on this norm index measurement—controlling for military status and rank, gender, race, age, income, education, and political knowledge (Appendix page 21). As expected, the norm index is a positive and significant predictor of both the out-party feeling thermometer (p < .04) and associating positive traits with opposing partisans (p < .01). In short, individuals with higher support for the apolitical norm hold less animus toward the other partisan side.

We have also yet to discuss differences between current and former service members. Recall that we designed our sample such that it was almost equally divided by military status (47.35% active-duty). A few things are noteworthy about military status. In regression models using the full sample (with controls for experimental condition and other variables), being active-duty (relative to veteran) is a significant predictor of two of the three measures of the norm that were successfully manipulated by our treatment (p < .05). People currently in the military have higher levels of support for the norm. Active-duty status is also a significant and positive predictor of the out-party feeling thermometer (p < .04) and associating positive traits with the opposing party (p < .01). All of this suggests that people currently serving in the military are less affectively polarized than veterans.

We might then ask whether military status (veteran/active-duty) moderates treatment effects. It does not appear to do so. Regressions interacting military status with experimental condition for the out-party feeling thermometer and positive trait rating reveal no significant heterogeneous treatment effects based on whether a person is active-duty or a veteran (Appendix page 22). As with the other tests in this section, these models include controls because these two groups differ in many respects that may impact these attitudes (e.g., age). We view this as a promising result because it suggests that the norm is a powerful tool for addressing affective polarization among service members even if they are no longer actively serving. In similar analyses in the Appendix (page 23), we test whether the treatment effects are moderated by military rank (officer/enlistee) and find null effects.

While we did not have any a priori reason to theorize asymmetric treatment effects by party, we performed regressions to test for this possibility. The results are in the Appendix (page 24). Neither treatment effect is significantly moderated by party. Nonetheless, our study was designed to test for average treatment effects, and a larger sample with more statistical power could detect heterogeneous effects where we do not.

Discussion and Conclusion

By some measures, hostility and discrimination toward partisan outgroups rival racial animus (Iyengar and Westwood 2015). These scholars suggest that racial discrimination is constrained by “egalitarian norms,” but “no such constraints apply to evaluations of partisan groups” (704). We argue and show that priming a group-based civic norm which includes an obligation to others cuts into affective partisan divisions.

Across feeling thermometers, trait ratings, and evaluations of the difficulty of serving with people of the opposing party, military service members appear affectively polarized. Although our survey was fielded during a time in which partisan divisions were particularly salient, ANES survey data shows that affective polarization among service members has been increasing for several years. The political views of people who have served in the military appear subject to the dynamics that have heightened partisan divisions among the broader population. This is important and merits attention because affective polarization may have consequences for decision-making among service members. Indeed, our party cue experiment reveals that service members think it is more difficult to serve with people from the opposing party.

Yet, there are reasons for optimism. Active-duty service members are, as expected, more likely to support the apolitical norm than veterans. People currently serving think it is important that the military rise above the partisan fray of contemporary politics. Beyond this, current service members, when compared to veterans, are more favorable in thermometer ratings of the out-party and more strongly associate positive traits with opposing partisans. It seems clear that people serving in the military—although affectively polarized—are less divided than their veteran counterparts.

Most importantly, we find evidence that priming an apolitical norm with an obligation to others lessens partisan divisions. Activating the norm increases positive feelings toward the out-party and the degree to which people associate positive traits with opposing partisans. Notably, military status (veteran/active-duty) does not moderate these treatment effects. This is reassuring in that it suggests that elevating the norm has effects for people even after they leave the service. Further, the norm moderates the influence of party cues on how difficult military service members believe it is to serve alongside and under command from the opposing party. There is affective polarization, but there is also a way to help address it.

Building on prior work, we focus on a norm with an “obligation to others” and the context of politics and partisanship, and we tailor the language of the norm to our audience. This study is a “proof of concept” about how group norms invoking an obligation to others can be used to combat affective polarization. Norms that entail an obligation to others are not unique to the military. Previous research with broader samples finds that civic norms with an obligation to others increase open-minded information search (Kam 2007), mitigate the influence of party cues on issue attitudes, and increase people’s willingness to discuss politics with a mix of partisans (Mullinix 2018). Given there is evidence that this type of norm impacts related outcomes with other samples, it seems plausible that the same strategy could be applied to other groups in the mass public. For example, if the target audience was evangelical Christians, researchers might prime a norm focused on an obligation to others such as, “Love your neighbor as yourself,'" and provide a suggestion (or sermon) about the implications for attitudes toward people across the partisan aisle. Just as we invoked military source cues our target audience viewed as credible, the quote could be from prominent evangelical Christians. Our study put the focus of the obligation to “others” on other “Americans.” There may be utility in broadening that to invoke a norm focused on an obligation to “humanity.” It is possible that this could not only temper affective partisan divisions, but also reduce hostility toward immigrants and refugees perceived as outgroups. We did not theorize about partisan asymmetries in responsiveness to our norm (nor find evidence of this), but it is possible that different norms resonate to varying degrees with Democrats and Republicans. As such, efforts to test the generalizability of our theoretical argument should recognize that the content and substance of civic norms may vary between groups. Researchers should consider the substance of a norm among their target population, look to previous literature and theory, and derive testable hypotheses about the effects of priming a particular norm. The key is activating an obligation to others and placing this in the context of partisanship.

We also see potential in related but distinct civic and social norms to mitigate affective polarization in the broader population and specific target populations.¹² We do not think that ability of norms to constrain partisan discord is limited to this particular norm and this specific sample. There is promise in efforts like Wojcieszak et al. (2020) and their use of a social norm of “open-mindedness.” There are multiple dimensions underlying civic norms (Conover et al. 1991), and, as such, multiple opportunities to explore strategies for addressing affective polarization among the population more broadly. For example, priming a sense of autonomy could have implications for feelings toward opposing partisans. Discussing this norm, Dalton (2008) states, “the good citizen should participate in democratic deliberation and discuss politics with other citizens, and ideally understand the views of others” (79). Yet, no one has tested the effects of priming this autonomy for affective polarization and this norm may resonate with the broader public.

Social norms have long been used as a tool to combat other types of prejudice; we are arguing that scholarship should apply these theoretical insights to affective partisan divisions. Our treatment unites the descriptive and injunctive aspects of norms, but it could be beneficial to disentangle these dimensions to pinpoint the most effective parts of norms for political behavior. And, given that schools and communities help develop civic norms (Campbell 2006), perhaps there should be discussions of integrating civility norms into civic curricula. In short, we think there are a wide-range of opportunities to apply and test our theoretical argument among various subgroups and the broader population.

We hesitate to overemphasize null effects, but our results illuminate limitations to using a common ingroup identity to address affective polarization. It is possible that a larger sample implemented at a different time might detect effects. Priming this identity may also be more or less impactful for divisions among different subsets of the population. The salience of national identity varies across demographics like immigration generation and race (Huddy and Khatib 2007), but without targeted oversamples of particular subgroups, we know little about the consequences of this variation for political behavior.

Affective polarization has risen in the United States. The extent to which it is problematic for civil society and governance is unclear, but it is concerning. As research continues to document how partisan hostilities permeate into our homes, workplaces, and classrooms, it is increasingly critical to identify strategies that effectively address affective polarization.

Footnotes

Acknowledgments

The authors thank Mark Joslyn, Jamie Druckman, Matthew Levendusky, and Matthew Miles for their feedback on earlier versions of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the University of Kansas

ORCID iD

Kevin J. Mullini

Notes

Author Biographies

Kevin J. Mullinix is an Associate Professor in the Department of Political Science at the University of Kansas. His research focuses on public opinion and political behavior in the United States.

Trent J. Lythgoe is a PhD candidate in the Department of Political Science at the University of Kansas and an Assistant Professor at the U.S. Army Command and General Staff College. His research focuses on American civil-military relations and political behavior in the U.S. military.

References

Abramowitz

2010. The Disappearing Center: Engaged Citizens, Polarization, and American Democracy. New Haven, CT: Yale University Press.

Abramowitz

A. I.

Webster

. 2016. The Rise of Negative Partisanship and the Nationalization of U.S. Elections in the 21st Century. Electoral Studies 41: 12‐22.

Ahler

D. J.

Sood

. 2018. “The Parties in Our Heads: Misperceptions About Party Composition and Their Consequences.” The Journal of Politics 80 (3): 964‐981.

Betros

2001. Political Partisanship and the Military Ethic in America. Armed Forces & Society 27 (4): 501‐523.

Blais

2000. To Vote or Not to Vote? Pittsburgh, PA: University of Pittsburgh Press.

Campbell

David E

. 2006. Why We Vote: How Schools and Communities Shape our Civic Life. Princeton, NJ: Princeton University Press.

Cialdini

R. B.

RenoReno

R. R.

Kallgren

C. A.

. 1990. A Focus Theory of Normative Conduct: Recycling the Concept of Norms to Reduce Littering in Public Places. Journal of Personality and Social Psychology 58 (6): 1015‐1026.

Conover

P. J.

Crewe

I. M.

Searing

D. D.

. 1991. The Nature of Citizenship in the United States and Great Britain: Empirical Comments on Theoretical Themes. The Journal of Politics 53 (3): 800‐832.

Crandall

C. S.

Eshleman

O'Brien

. 2002. Social Norms and the Expression and Suppression of Prejudice: The Struggle for Internalization. Journal of Personality and Social Psychology 82 (3): 359‐378.

10.

Dafoe

Zhang

Caughey

. 2018. Information Equivalence in Survey Experiments. Political Analysis 26 (4): 399‐416.

11.

Dalton

R. J

. 2008. Citizenship Norms and the Expansion of Political Participation. Political Studies 56: 76‐98.

12.

Dempsey

Jason K

. 2009. Our Army: Soldiers, Politics, and American Civil-Military Relations. Princeton, NJ: Princeton University Press.

13.

Druckman

J. N

Levendusky

M. S.

. 2019. What do we Measure when we Measure Affective Polarization? Public Opinion Quarterly 83 (1): 114‐122.

14.

Esper

Mark T

. 2020. Ethical Conduct and Political Activities. https://media.defense.gov/2020/Feb/06/2002245021/-1/-1/1/ETHICAL-CONDUCT-AND-POLITICAL-ACTIVITIES.PDF

15.

Feaver

Kohn

Richard H.

. 2001. Soldiers and Civilians: The Civil-Military Gap and American National Security. Cambridge, MA: MIT Press.

16.

Fiorina

M. P.

Abrams

S. A.

Pope

J. C.

Pope

Jeremy C.

. 2008. Polarization in the American Public: Misconceptions and Misreadings. The Journal of Politics 70 (2): 556‐560.

17.

Fiske

S. T.

1980. Attention and Weight in Person Perception: The Impact of Negative and Extreme Behavior. Journal of Personality and Social Psychology 38 (6): 889‐906.

18.

Garamone

2016. Active-Duty Personnel Must Remain Apolitical, Nonpartisan, Dunford Says. VA: U.S. Department of Defense. https://www.defense.gov/Explore/News/Article/Article/881624/active-duty-personnel-must-remain-apolitical-nonpartisan-dunford-says/

19.

Griffiths

Zachary

Simon

Olivia

. 2019. Not Putting Their Money Where Their Mouth Is. December: Armed Forces & Society. https://doi.org/10.1177/0095327X19889982

20.

Groenendyk

Eric W.

2013. Competing Motives in the Partisan Mind. New York: Oxford University Press.

21.

Holsti

O. R.

2001. Politicization of the United States Military: Crisis or Tempest in a Teapot? International Journal 57 (1): 1‐18.

22.

Huddy

Khatib

. 2007. American Patriotism, National Identity, and Political Involvement. American Journal of Political Science 51 (1): 63‐77.

23.

Huntington

Samuel P.

1957. The Soldier and the State: The Theory and Politics of Civil-Military Relations. Cambridge, MA: Harvard University.

24.

Iyengar

Krupenkin

. 2018. The Strengthening of Partisan Affect. Political Psychology 39 (Suppl. 1): 201‐218.

25.

Iyengar

Shanto

Lelkes

Yphtach

Levendusky

Matthew

Malhotra

Neil

Westwood

Sean J.

. 2019. The Origins and Consequences of Affective Polarization in the United States. Annual Review of Political Science 22 (7): 1‐18.

26.

Iyengar

Sood

Lelkes

. 2012. Affect, Not Ideology. Public Opinion Quarterly 76 (3): 405‐431.

27.

Iyengar

Westwood

S. J.

. 2015. Fear and Loathing Across Party Lines: New Evidence on Group Polarization. American Journal of Political Science 59 (3): 690‐707.

28.

Kam

C. D.

2007. When Duty Calls, Do Citizens Answer? The Journal of Politics 69 (1): 17‐29.

29.

Klar

Krupnikov

Ryan

J. B.

. 2018. Affective Polarization or Partisan Disdain? Public Opinion Quarterly 82 (2): 379‐390.

30.

Klein

J. G.

1991. Negativity Effects in Impression Formation: A Test in the Political Arena Personality and Social Psychology Bulletin 17 (4): 412‐418.

31.

Kohn

Richard.

2002. The Erosion of Civilian Control of the Military in the United States Today. Naval War College Review 55 (3): 8‐59.

32.

Lelkes

Westwood

S. J.

. 2017. The Limits of Partisan Prejudice. The Journal of Politics 79 (2): 485‐501.

33.

Levendusky

Matthew S.

2017. Americans, Not Partisans: Can Priming American National Identity Reduce Affective Polarization? Journal of Politics 80 (1): 59‐70.

34.

Levendusky

M. S.

2018. When Efforts to Depolarize the Electorate Fail. Public Opinion Quarterly 82 (3): 583‐592.

35.

McConnell

Margalit

Malhotra

Levendusky

. 2018. The Economic Consequences of Partisanship in a Polarized Era. American Journal of Political Science 62 (1): 5‐18.

36.

Monteith

M. J.

Deneen

N. E.

Tooman

G. D.

. 1996. The Effect of Social Norm Activation on the Expression of Opinions Concerning Gay Men and Blacks. Basic and Applied Social Psychology 18: 267‐288.

37.

Mullinix

K. J.

2018. Civic Duty and Political Preference Formation.” Political Research Quarterly 71 (1): 199‐214.

38.

Ohira

Winton

W. M.

Oyama

. 1998. Effects of Stimulus Valence on Recognition Memory and Endogenous Eyeblinks: Further Evidence for Positive-Negative Asymmetry. Personality and Social Psychology Bulletin 24 (9): 986‐993.

39.

Paluck

E. L.

2009. Reducing Intergroup Prejudice and Conflict Using the Media: A Field Experiment in Rwanda. Journal of Personality and Social Psychology 96 (3): 574‐587.

40.

Rogowski

J. C.

Sutherland

J. L.

. 2016. How Ideology Fuels Affective Polarization. Political Behavior 38 (2): 485‐508.

41.

Rohlfing

Ingo

. 2012. Case Studies and Causal Inference An Integrative Framework. New York, NY: Palgrave MacMillan.

42.

Shanahan

Patrick M

. 2019. Political Activities. https://news.usni.org/2019/06/11/acting-secdef-shanahan-reminds-pentagon-to-stay-apolitical-following-mccain-situation.

43.

Stangor

Sechrist

G. B.

Jost

J. T.

. 2001. Changing Racial Beliefs by Providing Consensus Information. Personality and Social Psychology Bulletin 27: 486‐496.

44.

Taber

C. S.

Lodge

. 2006. Motivated Skepticism in the Evaluation of Political Beliefs. American Journal of Political Science 50 (3): 755‐769.

45.

Tajfel

Henri

Turner

John

. 1979. An Integrative Theory of Intergroup Conflict. In The Social Psychology of Intergroup Relations, edited by Austin

W.G.

Worchel

Stephen

, Monterey, CA: Brooks/Cole.

46.

Tetlock

P. E.

1983. Accountability and Complexity of Thought. Journal of Personality and Social Psychology 45: 74‐83.

47.

Urben

H. A.

2014. Wearing Politics on Their Sleeves? Armed Forces & Society 40 (3): 568‐591.

48.

US Department of Defense . 2008. Directive 1344.10 Political Activities by Members of the Armed Forces. Washington, DC: US Department of Defense.

49.

US Department of Defense . Demographics Report. Washington, D.C.: Government Printing Office, 2018.

50.

Wojcieszak

Magdalena

Winter

Stephan

Xudong

. 2020. Social Norms and Selectivity. Mass Communication and Society.

51.

Young

H. P.

2015. The Evolution of Social Norms. Annual Review of Economics 7: 359‐387.