A Wedge Issue for the 21st Century: The Conditional Effect of Party Identification for Predicting Feelings Towards Immigrants and Refugees in a Higher Ed Setting

The issue of immigration played an important role in recent U.S. elections. How did the salience of immigration and refugees in the 2016, 2018, and 2020 elections play out amongst young voters? Were party identification and ideology central predictors of respondents’ feelings toward these groups, or were measures associated with social contact theory also predictive? The analyses here focus on a sample of college students, given that the higher education experience presents conditions for social contact theory to play out. We find that party identification is a key piece of the puzzle—party identification conditions how measures of social contact theory operates. These are important findings for those who wish to foster better intergroup relations in a higher ed setting and beyond.


INTRODUCTION
In 2016, Donald Trump ran a presidential campaign characterized by sharply negative messages about immigrants and refugees. Trump's framing of this issue dominated media coverage and had a powerful effect upon much of the electorate (Patterson 2016). Indeed, polls done just before the general election found that immigration was "very important" to seventy percent of likely voters, making it the sixth most important issue of the year (Pew Research Center 2016, 9). Trump supporters were particularly concerned, with seventy-nine percent indicating immigration was "very important" to their vote. Sixty-five percent of Clinton supporters agreed. However, with much of the rhetoric cooling by 2020, only fifty-two percent of likely voters deemed immigration to be "very important." The partisan gap remained, though, with sixty-one percent of self-identified Republicans denoting immigration as "very important" while only forty-six percent of self-identified Democrats felt the same. Even with this decline, immigration remained the eighth most important issue of the year (Pew Research Center 2020, 35-36).
Yet just because an issue such as immigration is deemed "important" does not mean individuals hold clear and unambiguous opinions about that issue. Ideology and partisanship are certainly crucial in understanding American attitudes toward immigrants and refugees (Chandler and Tsai 2001;Masuoka 2008). However, these variables can overshadow more nuanced explanations for an individual's beliefs, such as whether they have regular personal interactions with members of an outgroup, as well as the context (supportive or otherwise) in which these interactions take place. Consequently, the study of attitudes towards immigrants and refugees also draws upon intergroup contact theory for its analysis. This theory asserts that, under certain conditions, greater interactions between groups will lead to greater acceptance (Gimpel and Lay 2008). Members of one group may come into such interactions with a low tolerance for members of another group. However, as they spend more time with members of those other groups, their tolerance will increase in measurable ways (Singelman and Welch 1993).
Census figures indicate that twenty-two percent of the population in urban counties are foreign-born compared to just four percent in rural counties (Parker, Horowitz, Brown, Fry, Cohn, and Igielnik 2018). Consequently, people who reside in rural America tend to have more limited experiences with immigrants and refugees than those who live in urban and suburban communities (Fennelly and Federico 2008). However, Census data also show that the demographics of rural America are slowly diversifying, which implies that rural attitudes toward the foreign born may also be in flux.
For example, North Dakota is a relatively homogeneous state where much of its population live in rural areas 1 physically removed from the more diversified urban centers. The state appears to serve as a microcosm for the nation's conflicted feelings towards immigrants and refugees. While voters overwhelmingly supported Trump in 2016, subsequent protests against the Administration's immigration policies (Hagen 2018) as well as proposed changes in refugee resettlement procedures by the North Dakota state legislature (Hyatt 2017) indicate that the population did not share a clear consensus on such matters. In addition, the state's primary population centers, such as the cities of Fargo and Grand Forks, host significant numbers of immigrants and refugees and are steadily diversifying, much like the nation overall. 2 This study uses surveys of undergraduate students enrolled at North Dakota public colleges and universities during the 2016, 2018, and 2020 election cycles to examine attitudes towards immigrants and refugees. We construct regression models to determine which factors drove young adults' feelings towards immigrants and refugees. College students are in transition in many respects, developing their identities as adults, including their partisan identities. At the same time, transitioning to a college environment typically brings with it an expanded world, a chance to interact with new ideas and, potentially, new groups of people. With this sample of college students, we aim to assess, in a politically polarized environment, which factors will most successfully predict attitudes towards immigrants and refugees-will the heated and polarized political rhetoric provide the strongest predictors? Or might the increased opportunity to interact with a more racially and ethnically diverse population exert its own effect on these attitudeseither in positive or negative ways?

LITERATURE REVIEW
The most consistent predictors of a person's political attitudes are ideology and partisanship. While these two variables overlap to some degree, they should be treated as separate and distinct when considering such matters as immigration and refugee resettlement. "Most people in modern societies hold certain political beliefs, predispositions, or values which influence the positions they take on issues such as immigration policies" (Chandler and Tsai 2001, 179). This collection of attitudes serves as the basis of a person's ideology. While the majority are not ideologues or even particularly well-informed on the issues, "...most individuals possess such predispositions, often as a result of political socialization," their own self-interest, and group identification (179). Studies tend to operationalize ideology in terms of the left-right continuum (Chandler and Tsai 2001;Saxton and Benson 2003;Frasure-Yokley and Wilcox-Archuleta 2019) or the possession of certain cultural values (Haubert and Fussell 2006;Araujo, et al 2018). However defined, such measures generally have a strong relationship to an individual's attitudes, with conservative respondents having significant negative feelings towards outgroups while liberal respondents hold more positive feelings (de Vries, Hakhverdian, and Lancee 2013;Walker 2014;Mierina and Korojeva 2015).
Partisanship can be defined as the strength of an individual's support for a specific political party or organization. Studies of American attitudes towards immigrants typically operationalize this variable in terms of a scale ranging from strong Republican to strong Democrat (Neiman, Johnson, and Bowler 2006;Knoll, Redlawsk, and Sanborn 2011;Sanderson, Semyonov, and Gorodzeisky 2021). This research indicates that while Republicans and Democrats share a number of concerns regarding immigration policy, respondents who identify as strong Republicans tend to have more negative attitudes towards immigrants than those who identify as strong Democrats.
Despite any overlap, ideology and partisanship are not interchangeable as predictors of individual beliefs. Whereas ideology has proven to be a consistent predictor of attitudes towards immigrants, party identification can be a less reliable indicator. In some cases, the influence of party identification disappears when ideology is controlled (Fennelly and Federico 2008), suggesting ideology drives the relationship more than party identification.
The central question, then, is whether other forces that tend to shape individuals' attitudes toward outgroups even stand a chance against these partisan and ideological cues. A dominant theory in the field of intergroup relations, intergroup contact theory, might serve as a countervailing force for citizens' attitudes towards immigrants and refugees. In short, might those individuals who have more opportunity to interact positively with immigrant and refugee groups have more positive attitudes towards these groups, independent of the heated political rhetoric?
Social contact theory provides an additional explanation for citizens' attitudes towards immigrants and refugees. This approach has its roots in the work of Gordon Allport (1954), who argued that continuing interactions between members of antagonistic groups can reduce overall prejudice. However, intergroup contact by itself is not sufficient to bring about measurable change. Instead, Allport claimed, a reduction in prejudice requires four pre-conditions. These include: (1) participants in these interactions must treat one another as relatively equal; (2) groups must share common goals that require ongoing cooperation; (3) groups must embrace cooperation rather than competition to achieve those goals; and (4) the effect of such cooperation is "...greatly enhanced if sanctioned by institutional supports (i.e., by law, custom or local atmosphere)" (281). However, prejudices are likely to be inflamed if intergroup contacts occur under competitive conditions or situations where the minority group is perceived as impinging on the relative power of the dominant group.
Allport's work transformed the field, setting "into motion decades of research assessing whether and under what conditions intergroup contact diminishes hostility towards outgroups" (Paluck, Green, and Green 2019). This led to hundreds of quantitative and qualitative studies throughout the social sciences, though without reaching a clear consensus on when contact works to decrease outgroup prejudice. Seeking to resolve this issue, Pettigrew and Tropp (2006) derived a meta-analytic test of intergroup contact theory, using 713 independent samples from 515 studies. Whereas "past reviews of this vast literature have often reached conflicting conclusions regarding the likely effects of intergroup contact" (752), Pettigrew and Tropp were able to draw a set of coherent findings from their analysis. "The meta-analytic results clearly indicate that intergroup contact typically reduces intergroup prejudice" (766), thereby confirming the core concept of Allport's approach. The effects of intergroup contact "appears to be far broader than what many past commentators have thought." Indeed, intergroup contact had a positive impact on all groups under study, not just groups defined in terms of race or ethnicity. The policy implications were clear: intergroup contact has the "potential ... to be a practical, applied means of improving intergroup relations" (766), particularly in situations where there is institutional support for this interaction.
Updating Pettigrew and Tropp's evaluation, Paluck, Green, and Green (2019) performed their own metaanalysis with a focus on "contact studies that feature random assignment and delayed outcome measures, of which there were 27 in total" (129). They found that "the extent to which contact diminishes prejudice seems to vary according to the target of prejudice" (132). For example, intergroup contact worked well in reducing prejudice against individuals with disabilities, but less so for racial and ethnic groups. "It now appears that some types of prejudice may be more malleable than others, or that some combination of contact and prejudice mesh especially well" (153). Unfortunately, much of the research did not provide clear descriptions of the contact programs under assessment, thereby making it difficult to replicate those studies. "As a result, we learn little about what specific aspects of the contact are reducing participants' prejudice" (153). Zhou, Page-Gould, Aron, Moyer and Hewstone (2019) provided some sense of what was missing with their own meta-analysis, finding support for the extended contact hypothesis whereby "intergroup attitudes benefit from knowing that in-group members have cross-group friends" (154).
In contrast to these positive outcomes, the racial threat hypothesis posits that, under certain circumstances, interactions between members of antagonistic groups frequently can lead to even greater feelings of prejudice. If egalitarian attitudes are missing, the racial threat hypothesis predicts that contact between groups will lead to more intergroup prejudice. This result becomes even more likely if the dominant group believes it is losing longstanding advantages over the subordinate group (Blumer 1958). Key (1949) illustrated this dynamic in regard to relations between Blacks and Whites throughout the southern states, presenting evidence that Whites became more hostile as the number of Blacks within a community grew. An increasing Black population caused White voters to worry about an interracial competition for economic and political resources, which they feared was unwinnable over the long run (Blalock 1967). Scholars found support for this hypothesis in other regions of the United States (Huckfeldt and Kohlfeld 1989) as well as between Whites and other racial groups (Oliver and Wong 2003;Rocha and Espino 2009).
The conditions that determine whether intergroup contact will lead to more positive or negative attitudes towards outgroups are tied closely to local context. Here, we conceptualize local context as the difference among rural, urban, and suburban communities. Research has shown that this distinction can be a crucial factor in shaping attitudes towards immigrants and refugees. For example, people living in rural areas hold more negative attitudes than their urban and suburban counterparts, and this difference is attributable to rural residents': (1) minimal interactions with immigrants and minorities in general (Zarate and Shaw 2010); (2) greater worries about their economic future as well as a fear of entering into a competition for jobs with immigrants and refugees (Fennelly and Federico 2008); (3) ideological predispositions against immigration (Struthers and Bokemeier 2000); and, (4) lower levels of educational attainment (Burns and Gimpel, 2000;Hainmueller and Hopkins 2014). However, the research also suggests that rural attitudes are not monolithic as once supposed and have undergone significant revisions in recent years. While past generations of rural residents may have had limited contact with non-natives in their community and workplace, more recent generations are much more likely to have encountered immigrants and refugees in their public schools and elsewhere (Lay 2012). Indeed, Gimpel and Lay (2008) found that these contacts left rural youth much more tolerant of others than their parents and grandparents. In addition, Frasure-Yokley and Wilcox-Archuleta (2019) found that more subtle measures of community difference, such as tax capacity and whether respondents can be classified as evangelical, have significant impacts on how respondents form their attitudes towards immigrants. Thus, while rural citizens may have different experiences than those living elsewhere, their attitudes are not inherently negative and may even be in transition.
This study is primarily concerned with the rural versus urban status of the community in which respondents were living at the time of the survey. It is less interested in whether respondents originally came from rural or urban communities.

HYPOTHESES
The study tests six hypotheses drawn from the literature. Each of these predict how warm or cold an individual will feel towards refugees or immigrants, on a 100-point feeling thermometer, where 100 indicates that respondents feel very warm towards immigrants or refugees; 75 indicates that respondents feel warm towards immigrants or refugees; 50 indicates that respondents feel neither warm nor cold towards immigrants or refugees; 25 indicates that respondents feel cold towards immigrants or refugees; and 0 indicates that respondents feel very cold towards immigrants or refugees.
Hypothesis One: The stronger respondents identify with the Republican party, the colder their attitudes towards immigrants and refugees.
Research indicates that an individual's connection to a political party has a significant impact upon their attitudes towards immigrants and refugees (Tolbert and Hero 1996;Huber and Espenshade 1997;Neiman, Johnson, and Bowler 2006). This partisan difference has hardened and expanded up through 2018 (Sanderson, Semyonov, and Gorodzeisky 2021). Trump consistently emphasized negative attitudes towards immigrants and refugees-attitudes that were subsequently incorporated by a wide array of Republican by 2018 and 2020. At the same time, Democratic candidates mounted a concerted push back to these negative attitudes throughout the era, particularly post-2016. As a result, attitudes towards immigrants and refugees should be closely tied to party identification, with strong identifiers incorporating these attitudes more than weak identifiers and independents. Thus, we expect strong Republicans will have the coldest feelings towards immigrants and refugees while strong Democrats will have the warmest.
Hypothesis Two: The stronger respondents identify as conservative, the colder their attitudes towards immigrants and refugees.
Individual ideology has a powerful impact upon attitudes towards immigrants and refugees, though the results are not always straightforward. (Chandler and Tsai 2001). No matter their ideology, though, most individuals express some concern about the effects of immigration upon both society and government policy (Hetherington and Weiler 2009). However, individuals with more conservative ideologies tend to be more hostile towards immigrants and refugees. They also tend to emphasize immigration as an issue of great importance to American society (Knoll, Redlawsk, and Sanborn 2011). In contrast, liberal ideology is associated with a cosmopolitan worldview more accepting of non-native people (Espenshade and Hempstead 1996; Haubert and Fussell 2006).
Hypothesis Three: In the 2016 electoral cycle, the more attentive respondents are to the news, the colder their attitudes towards immigrants and refugees.
Above and beyond the effects of ideology and partisanship, we expect that an individual's propensity to pay attention to politics will shape their views on immigrants and refugees. However, this relationship is contingent on the electoral cycle. Whether liberal or conservative, voters who paid attention during the 2016 campaign would have had extended exposure to negative comments about these groups by nominee Donald Trump (Patterson 2016). While Democrats and Republicans get their information from different sources, it would be hard to miss the largely negative message about immigrants found in the media. Research from that election cycle indicates that the tone of coverage was overwhelmingly negative (Patterson 2016). Trump maintained his commitment to anti-immigration policies throughout his tenure, but the rhetoric surrounding the campaigns shifted over time. By 2018, Democrats were more focused on countering Trump's anti-immigration messages; and, by 2020, the national focus on the coronavirus obscured other issues. Consequently, we expect differing effects for those who report paying a lot of attention to the campaign-in 2016, we expect that, as attention increases, attitudes towards immigrants/refugees will become more negative. In 2018 and 2020, though, we expect this effect to fade away.
Hypothesis Four: Respondents who attend an urban campus have warmer feelings towards immigrants and refugees than respondents who do not attend an urban campus.
While many respondents come from rural areas, there is some variation as to whether respondents' daily lives resemble a more rural or urban environment-and, consequently, the probability that students might have the opportunity for cross-group interactions. The study is interested in where respondents currently live. In North Dakota, there are two campuses located in urban centers with relatively diverse populations. Students enrolled in these campuses, labeled "Urban Campus," have more opportunity to come into contact with immigrants and refugees in the context of institutional support. Consequently, students at these urban campuses will display more tolerant attitudes towards immigrants and refugees than students at non-urban campuses across the state.
Hypothesis Five: The greater the level of education, the warmer the feelings towards immigrants and refugees.
Greater educational attainment is typically related to a greater acceptance of immigrants and refugees. By focusing on college students, we can examine whether the effect of education is a selection effect-i.e., those who are already more positive towards immigrants and refugees select in to higher education, a culture that is known for positive attitudes towards these groups. Or if higher education itself is the root of more positive attitudes towards immigrants and refugees. Given the focus on undergraduate students, there is no variation in education levels as typically measured-all respondents have achieved a high school education but have not yet achieved a college education (though they are on track to do so). We do have variation in how long students have been immersed in a culture that is generally positive towards immigrants and refugees. If students become more positive towards immigrants and refugees as their time in higher education increases, then it is the culture that leads to these attitudes, and not a selection effect.
Hypothesis Six: The more negative the attitude towards the local economy, the colder the feelings towards immigrants and refugees.
The economic circumstances of individuals can be significant predictors of their attitudes towards immigrants and refugees. The research argues that native workers fear that immigrants and refugees will compete for jobs (Stephen, Ybarra, and Bachman 1999) or take advantage of government welfare programs, leading to higher taxes for the employed (Goldstein and Peters 2014). These be-liefs become even more pronounced when the economy is perceived as doing poorly. Following the racial threat approach, we predict that as negative attitudes about the state economy increase, negative attitudes towards immigrants/refugees will also increase.

DATA AND METHODOLOGY
This study uses survey data collected immediately after the national elections of 2016, 2018, and 2020. 3 For each round, we obtained a full contact list of students enrolled in North Dakota's eleven public colleges and universities. The morning after the election, students received an email invitation to participate in a survey about their political opinions. Of 45,359 students invited to participate in 2016, 4,810 completed the survey, for a response rate of 10.6 percent. Of the 45,745 students invited to participate in 2018, 2,678 completed the survey, for a response rate of 5.9 percent. In 2020, we invited 42,188 students to complete the survey; 3,432 did so, for a response rate of 8.1 percent. It should also be noted that, every two years, all enrolled students were sent an invitation to participate. While some students may have participated in the survey in more than one election year, the study does not identify or follow individual respondents from survey to survey. Finally, the surveys did not distinguish between respondents who were born in the United States and those who were born elsewhere.
North Dakota is an upper Midwest state with a population of 762,062 in 2019, up 13.3 percent from 2010 (US Bureau of the Census 2021). Some 33.3 percent live in the state's three largest cities: Fargo, Grand Forks, and Bismarck. The state's population is categorized as 86.9 percent white alone, 5.6 percent Native American, 4.1 percent Hispanic/Latino, and 3.4 percent African-American. These demographics are similar to other Midwestern states, such as Nebraska, Montana, and South Dakota. North Dakota's median household income was $64,894 income in 2020 with a poverty rate of 10.6 percent. The state's economy is largely driven by agriculture and oil, though higher education and tourism also play an important role.
While North Dakota shares similarities with other Midwestern states, it is quite different from the rest of the nation. These differences might cause some to question whether the results of such a single-state study are generalizable (Morgan, Hoyman, and McCall 2019). However, "...single state studies are appropriate when the researcher wishes to generalize to a unit of analysis other than the states themselves, when conditions in a given state provide a unique opportunity for the most rigorous test of a hypothesis, and when the measurement advantages of a single-state study outweigh the costs of limited generalization" (Nicholas-Crotty and Meier 2002, 411). The unit of analysis in this study is the individ- Students answered an array of questions related to their political behavior (e.g., whether or not they voted in the most recent election, how much attention they paid to the electoral campaigns); their political attitudes and opinions on a wide array of topics (e.g., the health of the state's economy); and key demographics. Table 1 displays the distribution of the sample for each electoral cycle on key variables. The distribution reveals relative stability across time. For example, when it comes to educational level, the samples are relatively evenly distributed across freshmen, sophomores, juniors, and seniors. 4 The sample consistently draws a higher percentage of women than men (from a low of 59 percent women in the 2018 sample to a high of almost 62 percent in the 2020 sample). The sample is also heavily white, with almost 96 percent of respondents in 2016 identifying as white and closer to 90 percent of respondents identifying as white in the 2018 sample. The sample is significantly whiter than the population. For example, in October 2020, the university system reported that just over 80 percent of enrolled students identified as white. 5 In terms of attitudes, Table 1 reveals a consistently conservative and Republican sample. On balance, the samples are optimistic about the state economy and very interested in electoral campaigns. Finally, the samples become progressively more positive towards immigrants and refugees over time. 6 In 2016, 8.7 percent of respondents reported very cold feelings towards refugees or immigrants, and 16.8 percent of respondents reported very warm feelings towards refugees or immigrants. By 2020, very cold feelings towards refugees or immigrants had dropped to 2.9 percent and very warm feelings towards refugees or immigrants had increased to almost 29.4 percent. Throughout all election cycles, the modal response to the immigrant and refugee feeling thermometer was "neither warm nor cold." Additional details of how these survey items were measured are available in Appendix A. Table 2 presents the results of a multiple regression model regressing the variables of interest on feelings towards immigrants and refugees across all three electoral cycles. We turn our attention to hypotheses 1 and 2, where we predict that party identification and ideology will explain feelings towards immigrants and refugees. These variables provide the strongest predictive power for these attitudes. Across all three electoral cycles, as a respondent becomes more strongly Democratic, their feelings towards refugees and immigrants become warmer. The average difference across all three models between a strong Republican and a strong Democrat is 21.9, a large difference on a 100-point scale. More specifically, in 2016, a strong Democrat, with mean values on all other independent variables, reports feelings towards immigrants and refugees at 68.3; a similar strong Republican in 2016 reports feelings towards immigrants and refugees at 45.3, slightly colder than neutral. 7 Similarly, in all three models, as a respondent becomes more conservative, their feelings towards refugees and immigrants become colder. While previous studies have found that the relationship between party identification and attitudes towards immigrants can disappear once ideology is accounted for, that is not the case in this sample; across all three models, party identification and ideology are both statistically significant. Given the Republican Party's standard bearer's strong emphasis on this issue, this is not surprising, and provides evidence that, at least in these three electoral cycles, the issue was fully incorporated into ideological and partisan identities. More specifically, across all three cycles, the average difference between strong liberals and strong conservatives is 24.4 points. Most pronounced is the difference between the average liberal and the average conservative in 2018: the average strong conservative indicated feelings towards immigrants and refugees at 49.1, while the average strong liberal indicated feelings towards immigrants and refugees at 83.0. This difference spans over one-third of the dependent variable's scale.

RESULTS
The third consistent finding is the effect of paying attention to campaigns on feelings towards immigrants and refugees-this effect, however, is more consistent than predicted. We anticipated that attention to the campaigns might reflect the elite consensus (or lack thereof) for each electoral cycle-for example, those who paid a lot of attention in the 2016 election would have been exposed, on balance, to negative messages about immigrants and refugees, potentially reflecting in the respondent's feelings towards those groups. We do not see evidence of this. Instead, across all three models, as attention to campaigns increases, feelings towards refugees and immigrants get warmer. This effect was relatively stable across all three election cycles; holding other independent variables at their means, the difference between those who reported paying very much attention to campaigns and those who paid not very much attention to campaigns was 8.1 points. The 2018 cycle fell the closest to this average. In that year, an average respondent who indicated they didn't pay very much attention to the campaign reported feelings towards immigrants and refugees at 50.4, the mid-point of the scale (i.e., feeling neither warm nor cold towards immigrants and refugees); those who indicated they paid very much attention to the campaigns felt warmer towards immigrants and refugees, with the mean level of 59.2. 8 In a test of the contact hypothesis, we turn to results for urban campuses (hypothesis 4). This variable provides a relatively consistent finding across two of the three models-students who attend urban campuses are statistically significantly warmer towards immigrants and refugees in 2016 and 2018, compared to students on non-urban campuses. This effect is much smaller than that associated with party identification or ideology. Holding all other variables at their means, an urban respondent in 2018 reports feelings towards immigrants and refugees at 60.8, as compared to nonurban campus respondents, who report an average value of 56.1. While there may be multiple mechanisms that lead to this relationship, the findings are consistent with contact theory's prediction that increased opportunity for interaction with immigrant and refugee populations will increase warm feelings towards these groups.
There is not consistent evidence for hypothesis fivethe prediction that attitudes towards immigrants and refugees will increase with year in school. We proposed that, if it was higher education itself that was causing warmer attitudes towards immigrants and refugees, then as a student's time in higher education increased, attitudes towards these groups should get warmer. We do not see this relationship consistently, however. Year in school only has a relationship to feelings towards immigrants and refugees in one electoral cycle (2016), where increased education leads to warmer feelings towards immigrants and refugees; otherwise, there is no relationship. Moreover, in analyses not shown here, this lack of relationship persists even when running the analysis on urban campus students only. This suggests that broader findings that those with more education have warmer feelings towards immigrants and refugees reflect a selection effect, in that those who already feel more positive opt for higher education.
In one final initial examination of the social contact theory, we interact urban campus with year in school (see Appendix B). This interaction can help account for the possibility of a selection effect for the urban campus findings-that is, that students who feel more warmly towards immigrants and refugees will opt into campuses where there may be more opportunity to interact with these individuals. If it is the urban campus experience that leads to warmer attitudes (and not vice versa), we should see a positive effect when we interact urban campus and year in school. In two out of three of these models (2016 and 2020), the interaction is insignificant and, in the third (2018), the interaction is statistically significant, but indicates a negative interactive effect (the year in school coefficient is 2.75 (SE = 1.12), the urban campus coefficient is 10.5 (SE = 3.25) and the interaction coefficient is -2.49 (SE = 1.25).) This suggests that any urban campus effects are attributable to selection effects.
Our indicator of the racial threat hypothesis is our measure of respondents' perceptions of the health of the state's economy. Once again, the results here are characterized by a lack of relationship-in only one of the three models do assessments of the state economy predict feel- ings towards the two groups (i.e., feelings towards immigrants and refugees in 2016). This relationship runs in the predicted direction-as assessments of the state economy increase, feelings towards refugees get warmer. However, the lack of consistency in this relationship does not provide robust evidence for hypothesis 6. Perhaps people only feel threatened when the state and national economy has truly experienced an economic downturn. North Dakota residents benefitted from a strong regional economy during the period under study here, thereby giving residents little reason to fear economic competition.

MIGHT PARTY IDENTIFICATION CONDITION SOCIAL CONTACT THEORY EFFECTS?
Thus far, the evidence suggests that political variablesparty identification, ideology, and attention to the campaigns-are the primary drivers behind attitudes towards immigrants and refugees. Indicators of the potential effects of social contact theory-i.e., whether an individual attends an urban campus and year in schoolappear to exert less consistent effects on the same attitudes. There is even less evidence for the racial threat hypothesis, in that perceptions of the state economy are only statistically significant in one of the three tested models.
Given the centrality of party identification, both in these models and in other work (e.g., Sanderson, et al 2021), it is worth considering whether there may be something about this identification that conditions the effects of social contact theory (and, by extension, the racial threat hypothesis). In particular, one of the key prerequisites for social contact to affect intergroup relationships in a positive way is that participants in these [potential and actual] interactions must treat one another as relatively equal (Allport 1954). Ideally, we would have a separate measure to assess individuals' views on the degree of equality between immigrants/refugees and the respondent's primary social identities. Without these specific measures, however, we can posit that as negative rhetoric and attitudes towards immigrants/refugees move towards the core of Republican identity, Republican identifiers will, on balance, be less likely to endorse egalitarian attitudes towards immigrants/refugees as a group. This attitude may be related to a lack of interactions and relationships with immigrants and refugees. As a result, then, we can expect to see the variables associated with social contact theoryi.e., year in school and attending an urban campus-be positive and statistically significant for Democratic iden-tifiers but not for Republican identifiers.
In addition, the findings in Table 2 for attentiveness to the campaigns do not support the original hypothesisi.e., that increased attentiveness should result in attitudes that mirror the dominant messages in the media. This would mean that, in 2016, when Trump's position on immigrants/refugees garnered disproportionate attention in the media, those who paid more attention to the campaigns should register more negative feelings towards immigrants and refugees. Similarly, this relationship might disappear, or even reverse itself, in 2018 and 2020 as Democrats pushed back more on this negative message and as other issues took center stage. Instead, the relationship between attention to campaigns and feelings towards immigrants and refugees is consistently positive.
Given the fragmented nature of the media environment, and citizens' increasing ability to seek out information sources that reinforce key political identities and attitudes, it is plausible that partisan identification might also be conditioning this relationship. As such, instead of expecting a message that is dominant in the aggregate to affect Republicans and Democrats in the same manner, we might expect instead that Democrats who pay more attention to the campaigns will seek out media sources that reinforce positive stances towards immigrants and refugees, leading attentive Democrats to have warmer feelings towards immigrants and refugees. On the other hand, Republicans who pay more attention to the campaigns will also seek out reinforcing media sources, consequently reinforcing (negative) party stances towards immigrants and refugees. This dynamic would lead attentive Republicans to have cooler feelings towards immigrants and refugees. Table 3 presents the results of a multiple regression analysis predicting feelings towards immigrants and refugees. This analysis breaks the sample into Republican and Democratic identifiers. Otherwise, the models are the same as before. The results replicate the predictive power of ideology-across all models, as respondents become more conservative, they feel colder towards immigrants and refugees. For example, in the most recent electoral cycle, holding all other variables at their means, a Republican who identifies as a strong conservative indicated feelings towards immigrants and refugees at 45.4, on the cold side of neutral; in that same electoral cycle, a Republican who identifies as an ideological moderate reports average feelings towards immigrants and refugees of 57.4. In that same cycle, an ideologically moderate Democrat indicates feelings towards immigrants and refugees of 69.4. Liberal Democrats in 2020, however, reported an average feelings of 93.6 towards immigrants and refugees. 9 We turn next to the variables that measure the effects of social contact theory. We do see party-conditioned effects here; however, in contrast with the party-conditioned hypothesis, social contact has a consistent and positive effect for Republican identifiers, and no effect for Democratic identifiers. In two out of three of the electoral cycles, as Republican identifiers progress through their college educations, they become warmer towards immigrants and refugees. For example, in the 2016 electoral cycle, a senior in college who identifies as a Republican reports a mean level of 46.3 for feelings about immigrants and refugees, as opposed to 40.9 for freshman Republicans. (This relationship did not hold for Republicans in the 2020 election.) These social contact theory variables are only statistically significant for Democratic identifiers, on the other hand, in the 2016 electoral cycle. Similarly, across all three election cycles, Republican identifiers on urban campuses were warmer towards immigrants and refugees than were Republican identifiers on non-urban campuses. (Urban campus Republicans across all three electoral cycles, when holding all other independent variables at their means, indicated a mean level of 48.6 towards refugees and immigrants, as compared to 43.7 for non-urban campus Republicans.) 10 Are there also interactive effects between party identification and attention to the campaigns?
Yes-Democratic identifiers become warmer towards immigrants and refugees as their attention to the campaign increases. This comports with the theory that greater attention leads to greater exposure and integration of the party's stance towards immigrants and refugees. This difference is particularly pronounced in 2016, where a Democrat who does not pay very much attention to the campaign reports average feelings towards immigrants and refugees at 59.9, as compared to 76.9 for Democrats who pay a lot of attention. For Republican identifiers, however, we do not observe the opposite relationship. Instead, there is no relationship, in any electoral cycle, between how much attention Republicans are paying to the campaign and feelings towards immigrants and refugees.
We didn't expect other interactive effects for party identification, and largely do not find them. Racial threat continues to exert very little influence for both Democrats and Republicans (though in 2016 Republicans do increase their positive feelings towards immigrants and refugees as their positive assessment of the state economy increases). In addition, the demographic control variables continue to show inconsistent effects, though Democratic women seem somewhat more positive towards immigrants and refugees than Democratic men in two electoral cycles (2016 and 2018).

CONCLUSION
We initially found that party identification and ideology were the strongest predictors of attitudes towards immigrants and refugees, while attentiveness to the news had a consistently positive impact on those attitudes, and social contact theory variables had inconsistent effects on those attitudes. The relationship between party identification and attitudes towards immigrants and refugees was so influential that we broke out the analysis by Democrats and Republicans, and these analysis bring some nuanced understanding to the relationships between our variables and attitudes towards immigrants and refugees. Party identification has become a central identity when it comes to attitudes towards refugees and immigrants.
Not only does partisan identity strongly predict these attitudes (as shown in Table 2), but it provides a framework for respondents such that variables matter differently for Republicans and Democrats. Republicans, in contrast to expectations, are more likely than Democrats to be affected by our social contact theory variables. This may be attributed to several mechanisms. For example, it may be that students who identify as Democrats are maxed out on positive attitudes towards immigrants and refugees. This ceiling effect would mean that, regardless of time in school or presence on an urban campus, Democrats feelings towards immigrants and refugees will be consistently high. If that is the case, then these findings suggest that social contact effects work for those who begin with the most negative attitudes.
The effects seen amongst Republican identifiers could also be the work of a selection effect. Especially for our urban campus indicator, it may be the case that Republicans with more positive feelings towards immigrants and refugees select in to urban campuses-and vice versa (i.e., that the Republicans with the most negative feelings towards refugees and immigrants opt for less diverse/non-urban campuses). As we discuss in endnote 10, this seems to be the case here. Further research could try to untangle what makes Republican (and Democratic) identifiers more amenable to selecting in to the more diverse urban campus environments.
On balance, these findings reinforce the findings of others that, while some common variables (i.e., party identification, ideology) can explain a lot about attitudes towards immigrants and refugees, there are key contextual variables-for example, what kind of community respondents inhabit-that also matter for these attitudes. What is more, party identification may be such a powerful variable for explaining these attitudes that Republicans and Democrats respond differently to these contextual variables. Such a moderated relationship would make sense in a world where feelings towards immigrants and refugees have become a central component of party polarization (Sanderson et al 2021).
Subsequent work has many avenues to pursue to better understand these nuanced relationships. Most obviously, given that we used a student population, do these effects hold for non-student populations or, for that matter, student populations in other states? Given the finding that the urban campus experience is associated with increased warm feelings towards immigrants and refugees for Republican identifiers, might those aspects of urban campuses that spark this relationship be replicated on non-urban campuses? And, finally, considering the constant evolution of the political context, will these particular attitudes continue to be at the core of partisan identities? If not, will these partisan differences in how attitudes towards immigrants and refugees are shaped continue once this issue becomes less salient?
The answers to these questions will continue to be important. Our campuses and society more broadly continue to search for ways to increase intergroup understanding. This study provides evidence that partisan identity may be another important contextual factor to take into account as we work towards progress in this area.
Dr. Nicholas Bauroth is a Professor of Political Science at North Dakota State University who specializes in state and local government, specifically special districts, local finances, and regional policies. His work has been published in State and Local Government Review, the American Review of Public Administration, the Journal of Urban Affairs, and other such periodicals.
Dr. Kjersten Nelson is a Professor of Political Science at North Dakota State University. She does work in American politics, specifically public opinion, gender, and the courts.

APPENDIX A
"Party identification" is modeled after the American National Election Studies survey that assesses strength of party identification with a seven-point scale where one indicates a strong Republican and seven is a strong Democrat. This was asked as a branching question. The percentage of respondents in our surveys who identified as "Strong Republican" ranged from 18.4 percent to 26.5 percent.
"Ideology" is modeled after the American National Election Studies survey that assesses strength of ideology with a seven-point scale where one indicates very liberal and seven is very conservative. Some 5.6 percent to 12.8 percent of respondents in our surveys identified as "Very Liberal" while 4.7 percent to 7.1 percent identified as "Very Conservative." "Attentive" is derived from the question "Some people don't pay much attention to political campaigns. How about you? Would you say that you were very much interested, somewhat interested, or not much interested in the political campaigns this year?" The percentage of respondents indicating that they are very much interested in political campaigns ranges from 48 percent (2020) to 32 percent (2018).
"Urban Campus" is defined as whether or not a respondent attends North Dakota State University (NDSU) in Fargo or the University of North Dakota (UND) in Grand Forks. Fargo is the largest city in the state while Grand Forks is the third largest. Both NDSU and UND are research universities with more than 13,000 students apiece while the other nine colleges and universities have fewer than 5,000 students each. A respondent who reports attending NDSU or UND receives a one while a respondent who reports attending some other state school receives a zero. The percentage of respondents who reported attending an urban campus ranged from 71.6 percent to 76.6 percent.
"Year in School" is defined as whether a respondent identifies as Freshman, Sophomore, Junior or Senior. Thus, the scale runs from one to four. Respondents who identified as Graduate or Professional are excluded from the analysis, as are those who identified as under 18 years old.
"North Dakota Economy" is derived using a Likert scale based upon the survey question "What do you think about the state of the economy these days in North Dakota?" Respondents can report Very Good, Good, Neither good nor bad, Bad, and Very Bad. The percentage of respondents who perceived the North Dakota economy as bad or very bad ranged from 41.5 percent to 55.0 percent.
There are three control variables in the model. These include: (1) "Woman" is defined as a binary variable based upon a respondent's self-identification. This variable was asked in an open-ended manner and coded as 1 for responses that indicated "female," "woman" or similarly feminine responses. It was coded 0 for responses that indicated "male," "man," or similarly masculine responses; (2) "White" is defined as a binary variable based upon a respondent's self-identification. Respondents could choose multiple categories and are coded as 1 if respondent chose "white" (and may have also indicated additional identities.) and (3) "Age" as reported by the respondent.
The full model is defined as: Attitudes = P artisanship + Ideology + Attentive to N ews and P olitics + U rban Campus + Y ear in School + N D Economy Bad + F emale + W hite + Age.

APPENDIX B
Appendix