Abstract
Recent evidence suggests that only about 1 in 5 U.S. adults has a friend on the political “other side” (Dunn, 2020). Although these interpartisan friendships are uncommon, they play a critical role in catalyzing empathy, reducing prejudice, furthering justice, and even restoring democracy, as suggested by the theory of civic friendship (Goering, 2003; Kahane, 1999; Rawlins, 2009). In the present study, we drew on national data from the Interfaith Diversity Experiences and Attitudes Longitudinal Survey (IDEALS) to examine the personal and contextual factors that predict interpartisan friendship formation among 5,762 college students attending 118 higher education institutions in the United States. The findings revealed the constellation of individual, social, and institutional contributors to students’ capacities to reach across political differences in their friendships. We offer guidance for how college educators can support the development of these relationships that may open a path toward empathy and healing in our polarized society.
Keywords
In his 2021 Memorial Day speech, President Biden declared that “he believes that ‘empathy is the fuel of democracy’ and asked Americans to ‘see each other not as enemies even when we disagree’” (Garcia, 2021, p. 1). Yet, popular rhetoric and news media coverage assert that political differences have become more deeply entrenched in the United States in recent years and that few are willing to reach across these boundaries to form relationships with those on the “other side.” Recognizing our deep-seated partisan divisions, sociologist Arlie Hochschild (2018) identified four steps that those within “liberal enclaves” need to take in order to “fix American democracy”; one step is “to reach out to people who grew up in geographic regions, classes, or religious groups very different from our own” (p. 1).
Recent evidence suggests that friendships across political differences (i.e., interpartisan friendships) are lamentably uncommon. A study by the Pew Research Center, conducted several months prior to the November 2020 U.S. presidential election, found that 39% of Trump supporters and 42% of Biden supporters reported having no friends who supported the other candidate, while only 22% of each group reported having at least some friends who supported the other candidate (Dunn, 2020). Pointing to higher education’s promising role in reducing political polarization, interpartisan friendships were somewhat more common among both Trump and Biden supporters with four-year college degrees (Dunn, 2020). Although friendship might be more likely among those who share political beliefs, there is nonetheless a robust minority of individuals who choose to form close relationships with those on the political “other side”—more than 1 in 5 U.S. adults, according to the Pew survey (Dunn, 2020). More research is needed to understand the factors that support the development of friendships that cross political boundaries—and what attending college may contribute to this process.
Literature Review
Our present study on college students’ interpartisan friendships, as well as our prior work on friendships across worldview differences (Hudson et al., 2021, 2022), is founded on the theory of civic friendship. Political theorists, including Aristotle (350 BCE/2019; Deneen, 2001), Tocqueville (Mallory, 2012), and Gadamer (1999; Walhof, 2006), emphasized that friendship does not take distinct political or personal forms; rather, within the context of a democratic society, all friendships serve some political function. Building upon the political functions of friendship identified by these earlier theorists, Kahane (1999) proposed the theory of civic friendship. In civic friendship, differences between friends foster personal growth by exposing individuals to “other stories” (Kahane, 1999, p. 285); our understanding of injustice becomes less abstract as we directly observe the effects of injustice on a person whom we care about, and this concern can then lead us to take political action on behalf of our friends (Goering, 2003; Rawlins, 2009). Furthermore, having a friend whose perspective and positionality differ from our own can provide the dissonance needed to help us rethink our own values and worldview. Indeed, friendships are unique among forms of interpersonal relationship because they can foster “moral transformation” (Friedman, 1993, p. 193) by challenging us to engage deeply with someone whose values may differ from our own; the greater the difference between friends, the greater the possibility for moral transformation.
The ideas underlying the theory of civic friendship highlight why friendships across social boundaries are a powerful way to enhance empathy and reduce prejudice across social divisions (Pettigrew, 1998a; Pettigrew & Tropp, 2000)—to not see those on the “other side” as enemies, as President Biden advocated in his 2021 Memorial Day speech. Yet, friendships are situated in a larger social context (Adams & Allan, 1998); in the United States, this context is characterized by deep-seated divisions along sociocultural boundaries. These patterns of social stratification, as well as stereotypes and assumptions about those “like us” (ingroups) and “different from us” (outgroups), shape our friendships by influencing with whom we come into contact and whom we perceive to be a potential friend (Vela-McConnell, 2011). This influence also runs in the opposite direction, however, with friendships having the potential to shape our social context; we can choose to form relationships that challenge the limitations that social stratification and divisions impose, congruent with the purposes of civic friendship. Friendship across social differences, therefore, “becomes a public statement that calls the stratification system into question” (Vela-McConnell, 2011, p. 21), modeling and making changes toward the more egalitarian and just society we wish to see (Rawlins, 2009; Vela-McConnell, 2011).
Within the specific context of higher education institutions, one of the most important educational and civic responsibilities “is to better position the next generation of leaders for the project of advancing social progress” (Hurtado, 2007, p. 186). Fostering the development of friendship across social boundaries during college—including the particularly intransigent boundary of partisanship—is one way to achieve this goal.
Partisanship as a Social Boundary
In the present study, we examine interpartisan friendships to identify the personal and contextual factors that facilitate students’ development of these relationships during college. In the U.S. context, the divide between those who identify as politically conservative or with the Republican party and those who identify as politically liberal or with the Democratic party is often termed partisanship. Partisanship operates as a social identity in ways similar to other social identities, such as race or religion, in terms of creating ingroup and outgroup dynamics within society; that is, we tend to view and treat others who share our partisan identity positively and those who do not negatively (Iyengar et al., 2019; Iyengar & Westwood, 2015; Kane et al., 2021; Wojcieszak & Warner, 2020).
Although there are intrapartisan differences in beliefs and values, as Egan (2020) observed, “for more and more Americans, politics has become key to the self-concept, leading ‘Democrat’ and ‘Republican’ as well as ‘liberal’ and ‘conservative’ to become identities in themselves that are meaningful far beyond shared policy preferences” (pp. 699–700). Strong dislike toward members of the opposing political group, rather than affiliation toward members of one’s own group, seems to drive partisanship behavior (Kane et al., 2021), further supporting the idea that partisanship operates as a formidable social boundary despite the presence of within-group differences among those who share a partisan identity.
That partisanship operates as a social identity is true for U.S. college students as well; while they may change positions on political issues during college, college students’ ideology and party affiliation (i.e., partisanship) have “strong affective components” and are “resistant to change” over four years (Woessner & Kelly-Woessner, 2020, p. 658). However, the political climate and norms on campus may lead some students—especially those whose political beliefs and partisan identities don’t align with the campus majority—to hide their partisanship or disengage politically to avoid being ostracized by peers (Curtis et al., 2019).
Conceptual Framework
Because the interpersonal dynamics of partisanship are similar to those for other social identities—that is, partisanship operates as a social boundary—it stands to reason that friendships across political (or partisan) differences may be similar to friendships across other social boundaries in terms of the factors that predict their formation. These predictors provide the conceptual framework for our study, guiding our choice of independent variables (which we discuss further in the Methods section). To organize our conceptual framework, we blended two similar conceptualizations of the levels of context influencing friendship: Adams and Allan’s (1998), which identified personal, network, community, and societal levels, and Vela-McConnell’s (2011), which specified individual, group, institutional, and sociocultural levels. In both conceptualizations, these levels of context influence the development of friendship as well as whether it is sustained over time. Given the purpose of the present study—to understand the contextual factors that facilitate students’ interpartisan friendships during college—our conceptual framework focuses on factors at the first three levels of context: community/institution, network/group, and personal/individual (see Figure 1). In the following discussion, we summarize the limited evidence regarding predictors of friendships across political differences and also review the evidence regarding predictors of friendships across social boundaries more broadly, organized by level of context, to provide support for our conceptual framework.
Conceptual framework and variable blocks included in analyses.
Community/Institution-Level Factors Associated With Friendships Across Social Boundaries
Institutional context has a significant influence on friendship because we most often meet potential friends through the social institutions to which we belong (Feld & Carter, 1998; Vela-McConnell, 2011) and because boundary-crossing friendships are more likely to occur in institutions that support members in overcoming social barriers (Vela-McConnell, 2011). Indeed, for college students, the community/institutional level of context may arguably be more salient than the societal/sociocultural context.
Prior research on college students’ friendships across social boundaries has highlighted how an institution’s climate for diversity and diverse perspectives may influence friendship. Rather than a single “diversity climate,” institutions have climates for various dimensions of diversity (e.g., race, worldview, political ideology). Racial climate has received the greatest scholarly attention to date regarding its influence on student development, interaction, and friendship. While a positive racial climate facilitates prosocial and other learning outcomes through the mechanism of interracial interactions (Chang et al., 2006; Jayakumar, 2008; Milem et al., 2005), the association between climate and interracial friendship has been mixed (e.g., Levin et al., 2003; Schofield et al., 2010). Similarly, while campus climate for worldview diversity has been found to positively predict students’ prosocial learning and development (Correia-Harker et al., 2019), a recent study of students’ friendships across worldview identity differences in the first year of college found no relationship between either a positive or negative campus climate for worldview diversity and interworldview friendship (Hudson et al., 2022). No research to date has examined campus climate as a possible predictor of friendship across political differences—insight that we seek to contribute with the present study.
Furthermore, students’ perceptions of campus diversity climates are shaped by their individual identities, which in turn may differentially influence whether students choose to engage with peers across social boundaries and how they experience those interactions (Chavous, 2005; Levin et al., 2003; Rolón-Dow et al., 2022). Recent research has also highlighted how institutional diversity, equity, and inclusion initiatives can fall short of fostering an inclusive climate for members of minoritized groups (Lewis & Shah, 2021; Rolón-Dow et al., 2022). When students feel unwelcome, tokenized, and forced to assimilate into the dominant culture, they understandably may be less likely to engage with peers who don’t share their identities and beliefs. As Curtis et al. (2019) found, this holds true for political beliefs as well.
An additional community/institution-level factor that may influence boundary-crossing friendship is the institutional type (i.e., private nonreligious, private religious, and public). In a previous study, we found evangelical institutional affiliation to be negatively associated with friendship across worldview identity differences (Hudson et al., 2022). Given the tight coupling between some religious denominations and political ideology in the United States and the related consequence that political diversity among students may, therefore, be lower at religiously affiliated institutions (making fewer politically different peers available as potential friends), we wish to explore possible associations between institutional type and interpartisan friendship. Additionally, Bowman and Park (2014) found public institutions to be negatively associated with interracial friendship among college students, providing further evidence for institutional type as a community/institution-level factor influencing boundary-crossing friendship.
Network/Group-Level Factors Associated With Friendships Across Social Boundaries
Our social networks and peer groups affect friendship development by influencing with whom we interact and whether we perceive another to be worthy of friendship (Adams & Allan, 1998; Feld & Carter, 1998; Vela-McConnell, 2011; Wimmer & Lewis, 2010). For example, racial homogeneity within our social networks means that our close friendships tend to be racially homogeneous as well because we tend to choose our friends from among those already within or adjacent to our existing social network (Wimmer & Lewis, 2010). Homogeneity in political ideology may operate in the same way, as suggested by Iyengar and Westwood (2015). Our perceptions of similarity to others may be influenced by the perceptions of peers or others whose opinions we trust (Selfhout et al., 2009; Sprecher, 1998), another way in which our social networks influence friendship development.
An additional factor at the network/group level concerns the influence of students’ social networks and peer groups on establishing and reinforcing campus norms regarding boundary-crossing interactions and friendship. Students who perceive boundary-crossing interactions to be common or expected on their campus are more likely to cross social boundaries to interact with peers (Chavous, 2005; Pettigrew, 1998b); thus, norms can create or constrain opportunities for development of friendships across social boundaries. Campus political norms also shape, and are shaped by, peer socialization; these norms may reduce students’ comfort in engaging with peers across political boundaries (Curtis et al., 2019).
Finally, limited prior evidence suggests disciplinary context (i.e., academic major) may shape boundary-crossing friendships at the network/group level. Hudson et al. (2022) found majoring in arts, humanities, or religion to be positively associated with, and majoring in social sciences or education to be negatively associated with, friendship across worldview identity differences. Regarding interracial friendship, Bowman and Park (2014) found a positive association (although only for some racial groups) with majoring in natural science or math/engineering and no association for other disciplinary contexts.
With the present study, we seek to provide insight into the associations between network/group-level factors, including disciplinary context and several forms of political engagement experiences with others on campus, and college students’ interpartisan friendships, as specified in the methods section.
Personal/Individual-Level Factors Associated With Friendships Across Social Boundaries
The very characteristic that makes friendships across social differences so transformative—the fact that they are boundary-disrupting—also makes them unlikely. At the personal/individual level, people gravitate toward relationships that affirm their closely held beliefs and values and find solace in perspectives and ways of life that are shared and familiar; this represents the idea of similarity-based attraction (Bahns et al., 2017; Huber & Malhotra, 2017; Johnson, 1989; Lizardo et al., 2015; Wimmer & Lewis, 2010). Similarity in political identity (or partisanship) has long been identified as a predictor of interpersonal attraction for adults generally and college students specifically (Iyengar et al., 2019; Lizardo et al., 2015; Rawlins, 2009; Rosenbaum, 1986). As Iyengar et al. (2019) note, partisanship “increasingly signals core values and worldview” (p. 136) and, therefore, functions as an important screening mechanism in selecting potential partners for friendship as well as other interpersonal relationships. For example, in a study of an online dating service, Huber and Malhotra (2017) experimentally manipulated dating profiles and then asked participants whether they would want to be friends with the person in the dating profile. They found that “shared ideology [i.e., self-identifying as liberal, conservative, or moderate] increases desire for friendship” (Huber & Malhotra, 2017, p. 276).
While deep-level similarity (e.g., attitudes, values, personality, interests) is important in initial attraction, it becomes less relevant within the friendship over time (Bahns et al., 2017), suggesting that if friends can find common ground beyond their political differences to form a friendship, that relationship may endure. One effective way to encourage college students to develop friendships across social boundaries is to provide ways for them to connect around shared interests and experiences (Hudson, 2018, 2022; Sacerdote & Marmaros, 2005).
Our individual experiences and attitudes are also components of the personal/individual level of context influencing friendship. Prior experiences with people of different identities, the quality of those experiences, and the level of comfort individuals have with those from backgrounds different than their own are especially relevant as personal/individual-level influences on boundary-crossing friendship by shaping our perceptions of and our expectations for interactions with others (Vela-McConnell, 2011). More specifically, whether an individual holds positive (appreciative) or negative (prejudicial) attitudes toward a particular outgroup may influence the likelihood of that individual seeking friendships with a member of that group—although the evidence is somewhat mixed. Among college students, two prior studies found a negative association between prejudiced attitudes toward other races at the start of college and interracial friendship (Bowman & Park, 2014; Fischer, 2008), whereas Schofield et al. (2010) found no association between two measures of racial attitudes (intergroup anxiety and prejudice) and interracial friendship. In a national, multi-institutional study of friendship across worldview differences among first-year college students, Hudson et al. (2021) found participants’ appreciative knowledge about other worldviews at the time of college entry not to be associated with the number of interworldview friends they reported at the end of the first year of college. Additionally, college students who report friendships across social boundaries in high school or at college entry are more likely to report boundary-crossing friendships during college (race: Bowman, 2012; Bowman & Park, 2014; Fischer, 2008; Park, 2014; Park & Kim, 2013; Stearns et al., 2009; worldview: Hudson et al., 2021). Such precollege friendship experiences may be both an indicator of students’ appreciative attitudes toward other sociocultural groups as well as a reflection of the prejudice-reducing effects of earlier boundary-crossing friendships shaping students’ behaviors in the college environment.
In general, members of majority groups have less experience interacting with members of marginalized groups; members of marginalized groups often have greater familiarity (if not always comfort) in interacting with majority-group members, which may make the development of boundary-crossing interactions and friendship more likely (Vela-McConnell, 2011). For example, Protestant college students, who represent a religious majority at many U.S. institutions, have lower levels of cross-racial interaction than Muslim, Buddhist, Jewish, and Hindu students (Park & Bowman, 2015). Regarding friendship, Stearns et al. (2009) conducted a single-institution study of how the racial heterogeneity of students’ friendship networks changed over the first year of college and found that friendship diversity increased for white students, decreased for Black students, and did not significantly change for Latino and Asian students; yet white students still had the most homogeneous friendship groups at the end of their first year of college compared to the other three racial groups. Additionally, students’ demographic characteristics complicate the relationships described previously between campus climate and interracial interaction and friendship. Notably, climate may be a stronger predictor of interaction for African American students than for other racial and ethnic groups (Chavous, 2005; Levin et al., 2003; Saenz et al., 2007). Thus, our analytic models contain three categories of independent variables at the personal/individual level of context: demographics, precollege experiences and attitudes, and partisanship.
Higher education holds tremendous potential for fostering boundary-crossing friendships by bringing together students of different social identities and views that span the political spectrum. Although supporting such friendships is vital to a successful democracy, college and university educators are left with little empirical guidance to drive this educational imperative. Most research on boundary-crossing friendships among college students has focused on friendships across racial (e.g., Bowman & Park, 2014; Fischer, 2008; Hudson, 2022; Martin et al., 2014) and, to a lesser extent, worldview identity differences (e.g., Hudson et al., 2021); little is known about friendship formation between college students of different political orientations. Given the capacity for friendships across political differences to reduce political polarization and increase empathy toward those on the “other side,” we sought to address the question: what contextual factors facilitate students’ friendships across political differences during college?
Methods
Data Source and Sample
The data source for this study, the Interfaith Diversity Experiences and Attitudes Longitudinal Survey (IDEALS), followed a national cohort of students across four years of college, centering their experiences with religious, secular, and spiritual diversity. Germane to this study’s research question, the survey included items pertaining to political orientation, voting, and political encounters with faculty and peers. Students were surveyed at three time points: the beginning (Time 1, fall 2015) and end (Time 2, spring/fall 2016) of their first year in college, as well as their fourth year (Time 3, spring 2019). IDEALS was conducted around the time of contentious presidential elections, which makes it an optimal resource for interrogating interpartisan friendships; the challenges of political polarization were particularly salient for college students during these years. Our analysis includes 118 campuses of different types (public, private nonsectarian, and religious) and represents all the major geographic regions in the United States. A total of 5,762 students participated in IDEALS Time 1 and Time 3.
Regarding characteristics of the sample, 54% identified as women, 46% as men, and 1% as another gender identity. Fifty-one percent of the sample identified as People of Color, including 20% who identified as Asian or Asian American, 13% as Latinx, 9% as multiracial, 7% as Black or African American, and 2% as another race or ethnicity other than white and including Native American. Nearly one-quarter (24%) were first-generation college students. The sample was made up of diverse religious, secular, and spiritual identities (RSSIs): 20% were another nonreligious RSSI, 18% were Catholic, 18% were evangelical Christians, 12% were mainline Protestants, 8% were atheists, 4% were another minoritized RSSI, 4% were another RSSI, 3% were another majoritized RSSI, 3% were Latter-day Saints, 3% were Muslim, 2% were Buddhist, 2% were Hindu, and 2% were Jewish. 1
Analysis and Variables
To address the study’s central research question, we performed a series of multilevel, multinomial logistic regressions that enabled us to examine a categorical dependent variable and effectively model the nesting of participants within institutions. Following listwise deletion, our analysis included 5,260 participants (91% of the sample), and we opted not to impute missing data given the low incidence of missingness (Allison, 2014). The dependent variable was generated on the basis of students’ reported political leaning in their fourth year of college (1 = very conservative to 5 = very liberal) and whether they had at least one friend who was very different from them politically (1 = yes, 0 = no). We categorized students into three groups: (1) “liberal” and “very liberal” students who made politically different (presumably conservative) friends (31%); (2) “conservative” and “very conservative” students who made politically different (presumably liberal) friends (10%); and (3) those who did not cross politically conservative-liberal boundaries in their friendships (58%). The third group included moderates and liberal or conservative students who did not cross political boundaries in their friendships. We calculated the relative risk, or likelihood, of identifying as a liberal with a conservative friend or identifying as a conservative with a liberal friend (compared to the baseline, those who did not cross political boundaries in their friendships) as a function of each variable in the model. A relative risk over 1.00 suggests that the likelihood of a particular outcome occurring is greater than that of the baseline outcome; correspondingly, a relative risk of less than 1.00 indicates that the baseline outcome is more likely to occur (Andrade, 2015). We weighted the data by gender (man/woman); race (white/Person of Color); type of school (private nonreligious, private religious, and public); Carnegie classification (doctoral program, masters, bachelors); geographic region (Northeast, Midwest, South, and West); and institutional urbanicity (city/not in a city) to approximate the college-going population attending four-year institutions in the United States. The data were also weighted to account for sample attrition over time.
The analytical models consisted of six categories of independent variables, which were grouped within the three major contextual layers featured in our framework (personal/individual, community/institutional, network/group; see Figure 1). Given the role of personal/individual contexts in friendships across differences (Adams & Allan, 1998; Bowman, 2012; Bowman & Park, 2014; Chavous, 2005; Feld & Carter, 1998; Fischer, 2008; Levin et al., 2003; Park & Bowman, 2015; Saenz et al., 2007; Stearns et al., 2009; Vela-McConnell, 2011), we included in the first category a series of dichotomous demographic variables, measured at the beginning of the first year on campus, reflecting gender identity; sexual orientation; race/ethnicity; first-generation status; and RSSI. The second category of variables, also grouped within the personal/individual contextual layer, reflected precollege experiences and attitudes as students began college (Bowman, 2012; Bowman & Park, 2014; Fischer, 2008; Hudson et al., 2021; Park, 2014; Park & Kim, 2013; Stearns et al., 2009)—namely, their attitudes toward politically liberal and conservative people (i.e., a scale reflecting the extent to which respondents agreed they had positive attitudes toward the group and agreed the group is ethical, makes positive societal contributions, and shares things in common with them), whether they had interpartisan friendships prior to college (i.e., dichotomous variables identifying liberals with conservative friends and conservatives with liberal friends), and the degree to which political views influenced their religious, secular, or spiritual worldview (0 = no influence to 3 = most influence). Capturing partisanship, voting patterns in the 2016 U.S. presidential election (i.e., dichotomous variables representing Trump voters and Clinton voters) made up the third category of variables included in personal/individual context. To address the community/institutional context (see Adams & Allan, 1998; Chang et al., 2006; Chavous, 2005; Correia-Harker et al., 2019; Hurtado et al., 2012; Jayakumar, 2008; Milem et al., 2005; Pettigrew, 1998b; Vela-McConnell, 2011), measures of institutional characteristics (i.e., institutional type/affiliation [public, private nonsectarian, Catholic, mainline Protestant, or evangelical Protestant] and students’ agreement that the campus welcomes politically liberal and conservative people [0 = disagree strongly to 5 = agree str-ongly]) constituted the fourth category of variables. The fifth and sixth categories of variables represented the network/group contextual influences on friendships across differences (Adams & Allan, 1998; Feld & Carter, 1998; Iyengar & Westwood, 2015; Selfhout et al., 2009; Sprecher, 1998; Vela-McConnell, 2011; Wimmer & Lewis, 2010). Students’ disciplinary context (i.e., arts and humanities; social science; health professions; science, technology, engineering, and mathematics; business; and double major) was included as a fifth category of dichotomous variables. Finally, the sixth category consisted of political engagement experiences with peers and faculty (i.e., taking courses in which faculty express liberal or conservative views [0 = never to 5 = all the time], feeling pressure to align political views with course instructors [0 = never to 5 = all the time], dedicating time to learning about politically conservative and liberal people [0 = disagree strongly to 5 = agree strongly], and having political disagreements with friends that result in either maintaining or ending the friendship [two dichotomous variables]).
All categorical variables with more than two categories (i.e., gender, sexual orientation, race/ethnicity, RSSI, institutional type/religious affiliation, and disciplinary context) were entered into the model as effect codes. Effect codes compare the mean for one group (e.g., students who identify as women) to the unweighted mean of all group means to reveal whether the group differs significantly from all others taken together (see Mayhew & Simonoff, 2015). This approach is preferred because it retains more information in the analytic models and avoids establishing an arbitrary (typically majoritized and privileged) reference group. Continuous variables were standardized. Table 1 provides descriptive statistics for all variables.
Descriptive Statistics
Results
Aligned with the frameworks of Adams and Allan (1998) and Vela-McConnell (2011), our results point to the relevance of personal/individual, network/group, and community/institutional levels of context in the development of collegiate interpartisan friendships (see Table 2). Across both models (conservatives with liberal friends and liberals with conservative friends), personal/individual qualities surfaced as meaningfully related to interpartisan friendships among students of different political persuasions. Specifically, politically conservative students who gain politically different friends in college are more likely than those who do not cross political boundaries in their friendships to identify as evangelical Christians (RR = 1.667, p < 0.05), report being conservative with interpartisan friendships at the start of college (RR = 2.162, p < 0.001), and have voted for Donald Trump in the 2016 presidential election (RR = 5.014, p < 0.001). Conversely, politically conservative boundary crossers are less likely than those who do not cross political boundaries to identify as Black or African American (RR = 0.351, p < 0.05); report being liberal with interpartisan friendships at the start of college (RR = 0.288, p < 0.05); indicate that political views shape their religious, secular, or spiritual identity (RR = 0.686, p < 0.05); and have voted for Hillary Clinton in the 2016 presidential election (RR = 0.383, p < 0.01).
Multilevel Multinomial Regression Models Predicting Political Boundary-Crossing Friendships
Note: RSSI is an acronym for Religious, Secular, and Spiritual Identity.
p < 0.05, **p < 0.01, ***p < 0.001.
Compared to their conservative counterparts, attitudes seem to play a more salient role in liberal students’ interpartisan friendship proclivities. That is, politically liberal students who make politically different friends in college are more inclined than those who do not reach across political boundaries in their friendships to report positive attitudes toward both liberals (RR = 1.273, p < 0.05) and conservatives (RR = 1.251, p < 0.001) at the beginning of college. Other factors predictive of interpartisan friendship among liberal-leaning students reveal parallels between liberals and conservatives. Like their conservative peers, liberals with interpartisan friendships have precollege experience with such friendships (RR = 2.673, p < 0.001) and tend to vote along party lines [(e.g., for Hillary Clinton in the 2016 presidential election (RR = 1.929, p < 0.001) and not for Donald Trump (RR = 0.134, p < 0.001)]. They are also less likely to identify as Black or African American (RR = 0.455, p < 0.001) than students who do not have interpartisan friendships. In other words, Black or African American students—regardless of political orientation—are less engaged in interpartisan friendships than students of other racial identities. By and large, personal/individual context contributed a greater array of factors to interpartisan friendship formation relative to other layers of context for students across the political spectrum.
With respect to the community/institutional context level, just one institutional characteristic reflective of campus climate surfaced. Conservative students with interpartisan friendships perceive their campus as welcoming of politically liberal people (RR = 1.401, p < 0.05). By contrast, campus climate did not definitively factor into the friendship patterns of liberal students.
Several factors in the group/network context are uniquely associated with conservatives who reach across political divides in their friendships: majoring in business (RR = 2.047, p < 0.05) and dedicating time to learning about politically conservative people (RR = 1.972, p < 0.001) increase the likelihood of interpartisan friendship formation while majoring in the arts and humanities (RR = 0.593, p < 0.05) and dedicating time to learning about politically liberal people (RR = 0.675, p < 0.05) are associated with a reduced likelihood of such friendships. Active efforts to sustain relationships through disagreement stand out as an important way for both conservatives (RR = 1.538, p < 0.05) and liberals (RR = 2.772, p < 0.001) to maintain their interpartisan friendships. Finally, politically liberal college students who feel pressured to align their political views with those of their course instructors are less inclined to cross political divides in their friendships (RR = 0.716, p < 0.001)—the same effect does not hold for politically conservative students, however.
We computed model fit statistics to compare the fit of models as each layer of context (i.e., personal/individual, community/institutional, and network/group) was added. Better-fitting models have higher log pseudolikelihood values and lower AIC and BIC values relative to other models. As shown in Table 3, the model fit generally improved as each successive contextual dimension was introduced. The log pseudolikelihood increases from Model 1 to Model 3, while AIC and BIC decrease (with one exception from Model 1 to Model 2 when BIC increases slightly). All told, Model 3, which includes all contextual layers, is the best-fitting model.
Model Fit
Discussion and Implications
A guiding assumption of this study is that friendships serve a civic purpose. Aligned with the theory of civic friendship (Kahane, 1999) and resting on the philosophical rationale for interpartisan friendships put forth by thought leaders across millennia (Aristotle, 350 BCE/2019; Deneen, 2001; Gadamer, 1999; Mallory, 2012), we contend that friendships that transcend political differences play a critical role in catalyzing empathy, reducing prejudice, furthering justice, and even restoring democracy (Friedman, 1993; Goering, 2003; Pettigrew, 1998a; Rawlins, 2009; Vela-McConnell, 2011). But to realize their potential outcomes, such “unlikely” (Vela-McConnell, 2011) friendships must be cultivated in the first place—and our study presents a framework for how interpartisan friendships come to be within higher education settings. Although not the majority, a considerable number of college students (41% of the sample in our study) reach across political divides in their friendships, which affirms that college students tend to be more successful at political boundary crossing than the general U.S. population (Dunn, 2020).
Our findings illuminate the key facilitators of interpartisan friendship formation, or friendships that disrupt the similarity-based attraction underscored in numerous studies (Bahns et al., 2017; Huber & Malhotra, 2017; Iyengar et al., 2019; Johnson, 1989; Lizardo et al., 2015; Rosenbaum, 1986; Wimmer & Lewis, 2010). On the whole, this study contributes a deeper understanding of how different contextual layers described by other scholars (Adams & Allan, 1998; Feld & Carter, 1998; Vela-McConnell, 2011), particularly the individual/personal, group/network, and institutional/community levels of context, function in relation to interpartisan friendship formation in the higher education environment.
First, our findings illustrate the powerful role of individual/personal context—identities, ideologies, and proclivities—in interpartisan friendship formation. Past research has shown that prejudicial attitudes can undermine friendships across social boundaries (Bowman & Park, 2014; Fischer, 2008), while exposure to such friendships prior to college can help (Bowman, 2012; Bowman & Park, 2014; Fischer, 2008; Hudson et al., 2021; Park, 2014; Park & Kim, 2013; Stearns et al., 2009). Further, identity (particularly race and religion or worldview) plays a meaningful role in interactions and friendships across social differences (Bowman, 2012; Bowman & Park, 2014; Chavous, 2005; Fischer, 2008; Levin et al., 2003; Park & Bowman, 2015; Saenz et al., 2007; Stearns et al., 2009; Vela-McConnell, 2011); people with minoritized identities often experience greater heterogeneity in their friendship networks than people with majoritized identities (Park & Bowman, 2015; Stearns et al., 2009).
This study offers further nuance while centering interpartisan friendships in particular. To begin, conservatives who establish friendships with liberals are more likely than their counterparts who do not cross political boundaries in their friendships to exhibit hallmark qualities of conservatism: they are evangelical Christians, 2016 Trump voters, business majors, and dedicated to learning about conservative people (and, by extension, they are less likely to be 2016 Clinton voters, arts and humanities majors, and dedicated to learning about liberal people). In parallel and signifying their political leanings, liberals who establish friendships with conservatives are more likely to be 2016 Clinton voters (and less likely to be 2016 Trump voters) compared to their peers who do not cross political boundaries in their friendships. In other words, conservatives and liberals can be who they are while they reach across the political divide in their friendships. Trump voters and Clinton voters do not need to forgo their presidential candidates of choice to be friends with one another. An important point about religious and political intersections is worth noting as well: conservatives with liberal friends are less likely than students without interpartisan friendships to report that politics shape their religious, secular, or spiritual identity. That they hold politics more loosely in relation to the domains of their inner life may open them to friendships across political differences. Arguably, a person who tightly couples political and religious ideologies might be more invested in politics as a meaning-making framework and remain wary of friendships that could potentially disrupt that framework (Wojcieszak & Warner, 2020).
Echoing past research (Bowman 2012; Bowman & Park, 2014; Fischer, 2008; Hudson et al., 2021; Park, 2014; Park & Kim, 2013; Stearns et al., 2009), attitudes and precollege experiences matter: experience with bridging political divides in friendships prior to college is predictive of such friendships later—among both conservatives and liberals. Moreover, liberals who begin college already holding positive attitudes toward both conservatives and liberals reach across political differences to form friendships with conservatives in college, a finding that reflects broader social patterns in which favorable views of the opposing party are associated with interpartisan friendship (Wojcieszak & Warner, 2020). Precollege experience with political boundary-crossing facilitates future relational success, which suggests that K–12 educational interventions are a worthy investment. Such interventions, which could become a robust component of civic education curricula, might entail educational activities that facilitate political self-understanding, considerate articulation of political beliefs, and skill-building for productive political exchanges and interpartisan friendships.
A final note about the role of the individual/personal context—in this case, race: this study provides an essential cautionary reminder that political boundary-crossing may present a hazard for minoritized populations. Black students, regardless of their political orientation, are less inclined to forge interpartisan friendships. And for good reason—as anti-Black racism persists, these relationships may pose too many risks. When white supremacy is interwoven with political ideology, friendships that attempt to bridge political divides are inherently unsafe for racially minoritized students. We suspect that safety concerns in interpartisan friendship may be a hurdle for students of other minoritized identities as well—queer and trans students, religiously minoritized students, and undocumented students, for example. When educators present opportunities for interpartisan exchange and relationship-building, conversations about personal boundaries and emotional safety are important. Sometimes, certain friendships are not possible; discussions that help students recognize the potential oppressions that come with friendships across differences should be incorporated into any training, classroom activities, or co-curricular opportunities featuring interpartisan friendship formation and sustainment. Students should be encouraged in these settings to identify what will work for them—and what will not—in their friendships so that their encounters are fulfilling and humanizing. Moreover, students with primarily privileged identities should be reminded that it is not the responsibility of their minoritized friends to serve as friend-educators—and that such undue burden is an injustice.
Plentiful evidence points to the influence of social or group networks on whether and how people develop friendships across social differences (Adams & Allan, 1998; Feld & Carter, 1998; Vela-McConnell, 2011; Wimmer & Lewis, 2010). Most of this work showcases the negative effects of homogeneous networks on friendships across difference (Iyengar & Westwood, 2015; Wimmer & Lewis, 2010) and how friends within our social networks, through their opinions and influence, affect our inclinations to make friends with outgroup members (Feld & Carter, 1998; Selfhout et al., 2009; Sprecher, 1998; Vela-McConnell, 2011). Our study adds further dimension to this line of inquiry, illuminating how exchanges within peer networks and experiences in classrooms affect interpartisan friendship formation. For example, both conservative and liberal students engaged in interpartisan friendships overcome the odds to establish such unlikely alliances—and they routinely navigate political disagreements to sustain their friendships. In fact, the salience of politics at the time IDEALS participants attended college meant that a majority (65%) of students had disagreements with friends about politics and remained friends nonetheless. That they had tenacity in the midst of deep disagreements appears to have helped their interpartisan friendships endure. At the classroom level, climates that were not politically coercive—where students did not feel pressured to align their political views with those of the instructor—facilitated liberal students’ friendships with conservatives. Faculty play an important role in setting the stage for interpartisan friendships—they model for students the possibility of productive dialogue and relationships when they open the classroom discourse to different perspectives. We are not of the mindset that faculty should keep all of their political viewpoints to themselves; a false neutrality may be read as inauthentic to their students. But if faculty choose to share where they stand politically, it is crucial that they simultaneously communicate their respect and goodwill toward people who have other opinions—as well as their openness to listening to and learning from others. A classroom environment where political differences are assumed and honored may go a long way toward helping students realize the potential for reaching across partisan differences and establishing friendships. That said, there are times when a line must be drawn between honoring political differences and enabling dehumanizing and harmful discourse. Faculty must take great care to intervene when political conversation moves toward undermining the humanity of any group or individual.
College educators and leaders can take from this study evidence that institution- and community-level factors matter. Politically conservative students with liberal friends attend campuses that they perceive as welcoming to liberals. Institutional norms and values are communicated through intentional efforts to welcome people across the political spectrum, and these norms appear to prompt conservative students to reach across political divides. Although campus climates for racial and worldview diversity foster prosocial outcomes generally (Chang et al., 2006; Correia-Harker et al., 2019; Jayakumar, 2008; Milem et al., 2005), the evidence for their impact on friendship formation across social boundaries is decidedly mixed (Hudson et al., 2022; Levin et al., 2003; Schofield et al., 2010). Moreover, the mere presence of people from different social identities—or the compositional diversity of the campus—is not always sufficient for developing prosocial outcomes and friendships across social boundaries (Bahns et al., 2011), especially when students perceive the campus climate to be unwelcoming or hostile to their identities and beliefs (Levin et al., 2003; Rolón-Dow et al., 2002). Our study affirms that particular aspects of campus climate matter the most and suggests that it’s important to characterize what we mean by “positive” and “negative” (Winkler et al., 2021) and move beyond assumptions that compositional diversity will support friendship formation on its own. The importance of norm-setting has been asserted by other scholars (Chavous, 2005; Levin et al., 2003; Pettigrew, 1998b), and our study shows these principles extend specifically to interpartisan friendships. In sum, a positive climate is a welcoming one—it’s a place where equity- and justice-centered institutional norms (like being welcoming) are clear to students and cue them to live within the expectations embraced by the campus community at large. Campus-wide mission and diversity statements, communications, and the actions of leaders convey norms and lead the way toward creating a welcoming environment where students of all political orientations are invited to engage and contribute.
This study has limitations that set the stage for future inquiry. Our focus here was on friendships that crossed the strong divide between political “sides” (i.e., partisanship), which is usually defined as a binary identity in the scholarly literature (e.g., Egan, 2020; Iyengar et al., 2019) as well as in national surveys and news media coverage, as we noted in the introduction. We assumed that students affirming that they had a politically different friend were referring to someone on the opposite side of the political spectrum (i.e., liberals who were friends with conservatives and conservatives who were friends with liberals). However, this framing assumes a simpler binary political system than the reality, and there may be more nuance behind students’ responses. Perhaps a student identifying as “very liberal” was thinking of their politically moderate friend; perhaps a Republican identifying as “conservative” was thinking of their Libertarian friend. Even though partisanship is well-documented as a powerful social identity divide, as discussed in our literature review, political beliefs are far more complex than what can be captured by the liberal/conservative (or Republican/Democrat) binary, and some college students choose to reject party affiliation entirely (Curtis et al., 2019). We encourage future researchers to capture multiple, more finely grained dimensions of political belief and ideology, ideally with multiple items that allow for intrapersonal nuance.
We were also limited in terms of the predictors of interpartisan friendships captured in our models. There are likely many other influences on friendships across differences than reflected in the variables we included. Future research that delves more deeply into the stories of college students’ political boundary-crossing—how they define “political difference,” the facilitators and motivating factors that help them build and sustain these relationships, and how other social identities interface with partisanship in these friendships—would help to advance our understanding of both the promise and perils these relationships introduce. Based on the distinctions this study revealed between liberals and conservatives regarding the facilitators of their interpartisan friendships, we surmise that college effects on boundary-crossing friendship formation could be conditional on student characteristics (e.g., political identity, gender, sexuality, race, religion). We urge researchers to explore these nuances more fully in the future. Despite the need for continued inquiry, this study is a step toward understanding and cultivating interpartisan friendships in college. The findings shed light on students who have managed to reach across political divides, illuminating who they are and how college supports their boundary-disrupting behavior—insight that may open a path toward empathy and healing in our polarized society.
Footnotes
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 Andrew W. Mellon Foundation under Grants 41600695 and 1712-05214; the Fetzer Institute under Collaboration Agreement 3816.00; and the Julian Grace Foundation under a grant dated July 1, 2017, to June 20, 2018 (no number assigned). The Interfaith Diversity Experiences and Attitudes Longitudinal Survey (IDEALS) was a collaboration between Interfaith America and principal investigators Matthew J. Mayhew (The Ohio State University) and Alyssa N. Rockenbach (North Carolina State University).
Notes
Authors
ALYSSA N. ROCKENBACH is alumni distinguished graduate professor at North Carolina State University, Department of Educational Leadership, Policy, and Human Development, Raleigh, NC 27695; email:
TARA D. HUDSON is associate professor of higher education at Kent State University, Department of Foundations, Leadership and Administration, Kent, OH 44242; email: