Attitudes about Immigrants and Refugees

Existing public opinion literature suggests that people’s attitudes about immigrants and refugees are typically not a matter of economic self-interest;Footnote ¹ rather, they are rooted in concerns about society and outgroup threat (e.g., Hainmueller and Hopkins Reference Hainmueller and Hopkins2014). Given such patterns, a psychological framework is useful for understanding variation in people’s attitudes about refugees. If concerns about the “other” characterize attitudes about immigrants and refugees, then it stands to reason that the psychological processes that determine whom people view as a potentially threatening outgroup are central to understanding this phenomenon.

Scholars have examined the links between psychological factors, such as values and emotions, and attitudes toward immigrants and refugees (Araújo et al. Reference Araújo, Bobowik, Vilar, Liu, Zuñiga, Kus-Harbord, Lebedeva and Gouveia2020; Brader et al. Reference Brader, Valentino and Suhay2008; Davidov et al. Reference Davidov, Meuleman, Billiet and Schmidt2008, Reference Davidov, Seddig, Gorodzeisky, Raijman, Schmidt and Semyonov2020; Paul Reference Paul2022; Seate and Mastro Reference Seate and Mastro2017; Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016; Verkuyten Reference Verkuyten2004). In response to reading about immigrants, people experience emotions such as anxiety (e.g., Brader et al. Reference Brader, Valentino and Suhay2008), contempt (e.g., Seate and Mastro Reference Seate and Mastro2017), anger (e.g., Verkuyten Reference Verkuyten2004), sympathy (e.g., Verkuyten Reference Verkuyten2004), and empathy (e.g., Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016). We engage with research that examines the pro-social effects of empathy on attitudes toward immigrants and refugees (Adida et al. Reference Adida, Lo and Platas2018; Simonovits et al. Reference Simonovits, Kézdi and Kardos2018; Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016; Sirin, Villalobos and Valentino Reference Sirin, Villalobos and Valentino2016; Williamson et al. Reference Williamson, Adida, Lo, Platas, Prather and Werfel2021). Extant work shows that treatments that trigger empathetic perspective-taking cause individuals to become more supportive of the immigrants and refugees they step into the shoes of (Simonovits et al. Reference Simonovits, Kézdi and Kardos2018; Adida et al. Reference Adida, Lo and Platas2018; Hartman and Morse Reference Hartman and Morse2018). For example, Adida et al. (Reference Adida, Lo and Platas2018) find that Americans prompted to do a perspective-taking exercise similar to those used by refugee advocates become more willing to write a letter to the White House on behalf of refugees. However, perspective-taking treatments are limited in their external validity (Sambanis and Simonson Reference Sambanis and Simonson2025).

Dissimilarity between the person making an evaluation about immigrants and the immigrants portrayed can change whether that person makes a positive or negative evaluation of immigrants (Bloom et al. Reference Bloom, Arikan and Courtemanche2015). For example, recent work suggests that racial/ethnic dynamics can shape the extent to which individuals empathize with immigrants in distress. After reading about an immigrant detainee’s treatment in a detention center, Sirin, Villalobos and Valentino (Reference Sirin, Villalobos and Valentino2016) show that, compared to whites, Blacks and Latinos have more group empathy for other racial/ethnic groups and thereby exhibit more supportive attitudes toward undocumented immigrants. Black respondents were more supportive of all detainees’ rights, regardless of their race, whereas white respondents were more supportive of white immigrant detainees compared to racial minority detainees. In a similar study using a measure of group empathy for other racial/ethnic groups, Sirin, Valentino and Villalobos (Reference Sirin, Valentino and Villalobos2016) show that racial minorities, especially Black Americans, have more empathetic responses and supportive policy attitudes toward immigrant detainees than white respondents. Taken together, these studies illustrate that people who are racial/ethnic minorities in the American context have higher levels of racial/ethnic outgroup empathy compared to the dominant racial/ethnic group (i.e., whites). We consider what these findings mean for public opinion about refugees, a group forced to migrate for humanitarian reasons. While refugees experience discrimination and disadvantage like undocumented immigrants, issues of legality are not central to their presence in the United States. Moreover, there is little reason to associate refugees with economic threat as their migration is humanitarian-based.

Existing work suggests that individuals view those who were forced to migrate (i.e., refugees) and voluntary economic immigrants in different ways. Associating migrants with the “refugee” label leads people to become more willing to help refugees and to express paternalistic stereotyping, while voluntary, economic migrants elicit opposition to helping and envy (Wyszynski et al. Reference Wyszynski, Guerra and Bierwiaczonek2020). Thus, existing evidence suggests that public opinion about refugees should be generally favorable, but does not uncover why Americans do not express similar levels of support for all refugee groups. We consider the role of race in American public opinion to address this puzzle.

Racial Attitudes and Public Opinion about Immigrant Groups

Racial identity and sentiment have long been central to American politics (Jardina and Piston Reference Jardina and Piston2019). Americans’ psychological attachments to their racial ingroups shape political attitudes and behaviors in powerful ways (Hutchings and Valentino Reference Hutchings and Valentino2004; Jardina Reference Jardina2021; Jardina and Piston Reference Jardina and Piston2019; Nava Reference Nava2023). Racial identity and racial attitudes influence policy attitudes (e.g., Hutchings and Valentino Reference Hutchings and Valentino2004), partisanship (e.g., Cuevas-Molina Reference Cuevas-Molina2023), support for civil liberty expression (e.g., Nava Reference Nava2023), and even conspiracy theory endorsement (e.g., Jardina and Traugott Reference Jardina and Traugott2019). For our purposes, we seek to link racial group dynamics with attitudes about refugees.

Race intersects with public opinion about immigrant groups and immigration in several key ways. Individuals exhibit racialized preferences in their immigration attitudes (e.g., Brader et al. Reference Brader, Valentino and Suhay2008). For example, people tend to prefer refugees and immigrants of the same ethnicity as most of the host population (Coninck Reference Coninck2020). Schachter (Reference Schachter2016) shows that white Americans are open to relationships with most immigrants, except for Black immigrants. Such work suggests that the race associated with refugees will influence the level of support expressed toward those refugees. Importantly, there is heterogeneity in the extent to which public opinion about immigration is racialized. The bulk of American public opinion about immigration focuses on white attitudes. As this scholarship has expanded to include more studies of the attitudes of racial minorities, it suggests that the foundational underpinnings of immigration may differ by racial group (e.g., Carter Reference Carter2019; Hutchings and Wong Reference Hutchings and Wong2014; Vienrich and Creighton Reference Vienrich and Creighton2018).

Preferences for one’s racial ingroup tend to be stronger than animus toward a racial outgroup (Tajfel and Turner Reference Tajfel and Turner2004), suggesting that preference for ingroup members outweighs one’s dislike of outgroup members (see Jardina (Reference Jardina2021) for a detailed discussion). Black Americans tend to report more supportive attitudes toward immigrants compared to white Americans (Brader et al. Reference Brader, Valentino, Jardina and Ryan2010; Kinder and Sanders Reference Kinder and Sanders1996), especially if the immigrants in question are racial/ethnic minorities (Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016; Sirin, Villalobos and Valentino Reference Sirin, Villalobos and Valentino2016). At the same time, Black Americans have been documented to express restrictive immigration attitudes (Carter Reference Carter2019; Nteta Reference Nteta2013) and hold more negative views of immigration compared to Hispanics (Vienrich and Creighton Reference Vienrich and Creighton2018).

We focus on the attitudes of Black and white Americans, following in the footsteps of research that studies these groups to better understand the relationship between race and immigration attitudes in American politics (e.g., Carter Reference Carter2019; Hutchings and Wong Reference Hutchings and Wong2014; Marrow et al. Reference Marrow, Tropp, van der Linden, Okamoto and Jones-Correa2019; Nteta Reference Nteta2013; Sheares Reference Sheares2023). We do so for the following reasons. First, Black and white identities are the most longstanding, politically salient racial identities among historically “native-born” groups in the United States (Hutchings and Wong Reference Hutchings and Wong2014). These groups are consistently at opposite ends of constructed societal hierarchies that uphold white supremacy. Blacks and whites also have a unique relationship in American politics with considerable systemic and formalized oppression against Black Americans that was not experienced in the same magnitude by other racial groups (Sellers et al. Reference Sellers, Smith, Shelton, Rowley and Chavous1998).

Historically, immigration has been highly salient for Black Americans. Immigration policy has had a direct impact on African Americans by shaping economic opportunities and signaling Black citizens’ status in American society (Diamond Reference Diamond1998; Hutchings and Wong Reference Hutchings and Wong2014; Carter Reference Carter2019). Scholars have argued that this positionality leads Black and white Americans to hold distinct frameworks for their immigration views (Bobo and Hutchings Reference Bobo and Hutchings1996; Carter Reference Carter2019; Diamond Reference Diamond1998; Hutchings and Wong Reference Hutchings and Wong2014; Nteta Reference Nteta2013). As Carter (Reference Carter2019) aptly argues in her book American While Black: “The fear of being further marginalized is really at the core of black engagements with immigration…” (p. 158). The literature suggests that White Americans’ attitudes about immigration are substantially racialized, while this does not appear to be the case for Black Americans (Carter Reference Carter2019; Ostfeld Reference Ostfeld2017; Marrow et al. Reference Marrow, Tropp, van der Linden, Okamoto and Jones-Correa2019; Sheares Reference Sheares2023). Black Americans consider both the group interests of African Americans, which can be threatened by increased immigration, and support for egalitarianism (Carter Reference Carter2019; Hutchings and Wong Reference Hutchings and Wong2014). We posit that refugees are less likely to signal group threat to Black individuals as their migration is humanitarian rather than economic in nature.

While we expect to observe a preference for racial ingroups among Black and white individuals, the strength of ingroup preference will likely be weaker for Black individuals compared to white individuals. Although Black and African Americans tend to favor African immigrants over others (Carter Reference Carter2019), they do not necessarily view each other as co-racial. While non-Black groups tend to view Black individuals as a homogeneous group that engages in politics in a united manner, Black individuals negotiate within-group diversity that can lead to divergences in opinion (Smith Reference Smith2014). Black immigrants diverge from U.S.-born Blacks in their attitudes on some policies, such as the belief that immigrants steal jobs from Americans and general preferences about the level of immigration in the United States (Capers and Smith Reference Capers and Smith2016; Unnever and Gabbidon Reference Unnever and Gabbidon2015). Linked fate leads to support for marginalized Black groups (e.g., undocumented immigrants), but does not apply to every aspect of Black politics (Bunyasi and Smith Reference Bunyasi and Smith2019).

Outgroup vs Ingroup Empathy

Empathy refers to the capacity for sharing and understanding the perspectives and feelings of others (Davis Reference Davis1983; Zaki Reference Zaki2014). Though it is a matter of some debate among scholars of empathy, the phenomenon contains cognitive, affective, and motivational components (Zaki Reference Zaki2014, Reference Zaki2017). To differentiate these components in simple terms, the cognitive dimension of empathy refers to an individual’s ability to take the perspective of another, the affective dimension refers to a person’s ability to feel or experience the state of another, and the motivational dimension refers to the desire for the improvement of the emotional state of another (Zaki Reference Zaki2014).

Empathy is a psychological trait that varies at the individual level (i.e., dispositional empathy) (Davis Reference Davis1983) and can be operationalized in reference to specific groups (i.e., group empathy). Group empathy measures the level of empathy a person exhibits for a particular group. Like interpersonal empathy, group empathy is a trait, or predisposition, that begins to develop in infancy, grows throughout adolescence and into middle-age adulthood, and moderately declines in old age (O’Brien et al. Reference O’Brien, Konrath, Grühn and Hagen2013; Szalavitz and Perry Reference Szalavitz and Perry2011). As Sirin et al. (Reference Sirin, Valentino and Villalobos2017) discuss, group empathy (as opposed to interpersonal empathy) is important for explaining gaps in political and policy attitudes. Independent from trait empathy, situations can elicit empathy (e.g., Batson et al. Reference Batson, Chang, Orr and Rowland2002). While this form of empathy has important implications for political behavior, we focus on the explanatory effects of trait group empathy.

Notably, empathy does not always lead to pro-social effects. In fact, empathy can exacerbate prejudice rather than reduce it (Brophy and Mullinix Reference Brophy and Mullinix2023; Simas et al. Reference Simas, Clifford and Kirkland2020). We explore the limitations of the pro-social effects of empathy on attitudes toward refugees by coalescing research on the racial underpinnings of political attitudes with social and developmental psychology’s work on group empathy. Experiencing empathy is a costly task for an individual and therefore people tend to “downregulate” empathy to reduce its cost (Kurzban et al. Reference Kurzban, Burton-Chellew and West2015). While scholars of empathy disagree on the extent to which empathy is socially biased (Fowler et al. Reference Fowler, Law and Gaesser2020), evidence suggests that people empathize more with perceived ingroup members than perceived outgroup members because it is an easier task (Behler and Berry Reference Behler and Berry2022; Bruneau et al. Reference Bruneau, Cikara and Saxe2015; Bruneau et al. Reference Bruneau, Cikara and Saxe2017; Cikara et al. Reference Cikara, Bruneau, Bavel and Saxe2014; Kurzban et al. Reference Kurzban, Burton-Chellew and West2015). The amount of empathy that individuals experience for others depends on the traits and circumstances associated with the person or group.

The results of such patterns are non-trivial. When ingroup empathy is stronger than outgroup empathy, individuals are willing to passively harm other groups (Cikara et al. Reference Cikara, Bruneau, Bavel and Saxe2014; Bruneau et al. Reference Bruneau, Cikara and Saxe2017; Zaki and Cikara Reference Zaki and Cikara2015). For example, Bruneau et al. (Reference Bruneau, Cikara and Saxe2017) examine the group dynamics between native-born Hungarians and Muslim refugees. They find that as Hungarians’ ingroup empathy increases, they become more likely to support anti-refugee policies, less likely to support increases in the number of Muslim refugees allowed to resettle in Hungary, and less likely to sign a pro-refugee petition. As outgroup empathy increases, the opposite pattern occurs. In the American context, racial minorities exhibit more group empathy for other racial/ethnic groups and this predicts their higher levels of support for immigrants of color compared to Anglos (Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016; Sirin, Villalobos and Valentino Reference Sirin, Villalobos and Valentino2016). Taken together, research on the effects of group empathy suggests that while higher levels of empathy are typically associated with cross-group support, this relationship can change depending on whether the object of empathy shares a perceived ingroup membership.

Data

To test our hypotheses, we conducted a web experimentFootnote ³ on a sample of 3,000 respondents from Prolific, a survey platform used by groups such as Stanford, Meta, and Kickstarter. Prolific is a high-quality survey provider with more representative and less professionalized samples than other cost-effective providers, and allows researchers to easily oversample specific demographic groups (Palan and Schitter Reference Palan and Schitter2018). Prolific uses population weights to collect representative samples of the population of interest. For this analysis, half of our respondents self-identify as Black and the other half as white. On other demographic factors, such as age, gender, and education level, our sample approaches population representativeness.Footnote ⁴ Due to group empathy’s reliance on a clear ingroup preference, we do not include mixed-race individuals, Hispanic whites, or Arab whites. For the Black subsample, we exclude mixed-race individuals as well as Hispanic and Arab Blacks for the same reasons. We also exclude individuals who failed simple attention checksFootnote ⁵  and those who took less than 2 minutes or more than 20 minutes to complete the survey due to likely noncompliance (Harden et al. Reference Harden, Sokhey and Runge2019). This results in a sample size of 1,353 white respondents and 1,332 Black respondents.

Measurement

Respondents were randomly assigned to a racially congruent or incongruent treatment with equal probability. Racial priming was used to cause respondents to consider a specific race while answering subsequent, seemingly unrelated questions (Mendelberg Reference Mendelberg2008). To maintain a balance between ensuring that the racial prime sufficiently elicits an effect while avoiding social desirability issues, we used an image-based prime. Extant literature shows that image primes have enough information for a respondent to determine race without issue (Abrajano et al. Reference Abrajano, Elmendorf and Quinn2018) and for respondents to determine if the individual is in their racial ingroup or not (Moehler and Conroy-Krutz Reference Moehler and Conroy-Krutz2016). Images without explicit, racialized language can still activate racial bias (Hutchings and Jardina Reference Hutchings and Jardina2009; Mendelberg Reference Mendelberg2008; Valentino et al. Reference Valentino, Hutchings and White2002; Weisbuch et al. Reference Weisbuch, Pauker and Ambady2009), rendering explicit racial appeals unnecessary.

In fact, existing evidence suggests that subtle racial manipulations are the most successful type of racial prime for eliciting racial considerations in web and laboratory experiments (Hutchings and Jardina Reference Hutchings and Jardina2009). The implicit nature of a subtle race cue is more effective in reducing bias activation compared to explicit cues (Dovidio et al. Reference Dovidio, Kawakami, Johnson, Johnson and Howard1997; Nteta et al. Reference Nteta, Lisi and Tarsi2016). Additional research has indicated that when a treatment is overtly racialized (even with a treatment that only involves imagery), the respondent may be triggered to engage in self-monitoring where they work to counteract their own negative stereotyping to maintain a positive sense of self or to reduce judgment from others (Dovidio et al. Reference Dovidio, Kawakami, Johnson, Johnson and Howard1997; Hutchings and Jardina Reference Hutchings and Jardina2009; Mendelberg Reference Mendelberg2008; Nteta et al. Reference Nteta, Lisi and Tarsi2016).

Using published news stories about refugees, we selected two images of a single refugee woman, one from Ukraine and the other from the Democratic Republic of the Congo. The woman from Ukraine is white and the woman from the Democratic Republic of the Congo is Black. The women in the photos appear to be similar in age, expression, and other attributes.Footnote ⁶ The background of these images has been blurred to reduce any extraneous information and to isolate the primary difference between the photos as the race of the individual shown ensuring proper internal validity. Respondents are told in the prime that the person in the photo is a refugee, but are not told which country the refugee is from or any other personal information. Respondents are given the definition of what it means to be a refugee that reflects the official definition according to the United Nations Human Rights Commission. As shown in Figure 2, the image is the only variation between treatments. The text-based prime remains constant. The combination of the image variation and the racially congruent or incongruent assignment results in four treatment groups (see Table 1).

Figure 2.

Racial prime treatments.

Table 1.

Treatment groups

After the respondent was shown the image with the definition of the term “refugee” they responded to a battery of questions to assess their group empathy for refugees. To measure group empathy for refugees, we utilize the Group Empathy Index (GEI).Footnote ⁷ The GEI adapts a well-established measure of interpersonal empathy, the Interpersonal Reactivity Index (IRI) (Davis Reference Davis1983). The IRI of one of the top three most commonly used measures of dispositional empathy and has been validated in several other languages (Gilet et al. Reference Gilet, Mella, Studer, Grühn and Labouvie-Vief2013; Lima and Osório Reference Lima and Osório2021).

While the IRI measures four subscales of empathy, perspective-taking, empathetic concern, personal distress, and fantasy, the GEI only includes measures for perspective-taking and empathetic concern since they are empirically linked to other-oriented behavior (e.g., Bobba and Crocetti Reference Bobba and Crocetti2022; Konrath Reference Konrath2013; Miklikowska Reference Miklikowska2018). The dimensions of empathy, typically called fantasy and personal distress, are not included as they tend to predict self-oriented outcomes rather than other-oriented ones (Konrath Reference Konrath2013). The GEI and similar measures of dispositional empathy rely on respondent self-reporting as opposed to neural imaging. While a survey-based response is more vulnerable to social desirability bias than a brain scan (Cikara and Fiske Reference Cikara and Fiske2011), the empirical patterns across measurement type typically align (e.g., Lamm et al. Reference Lamm, Batson and Decety2007; Mathur et al. Reference Mathur, Harada, Lipke and Chiao2010).

We adjust the GEI to measure group empathy for refugees.Footnote ⁸ We display the exact question wording in Table 2 and note which statements measure perspective-taking and which measure empathetic concern. Response options consist of a 6-point agreement Likert scale where higher values signify higher levels of group empathy. This variable, an additive index, has a range of 14–84 with a mean of 64. Figure 3 visualizes the distribution of this index.

Table 2.

Measure of group empathy for refugees

Figure 3.

Distribution of group empathy index.

Respondents were then asked a variety of demographic questions, such as their party identification, gender, religion, education level, and age, followed by a question determining their degree of linked fate (or belief that their fate is linked to the fate of others of their same race in the United States) (Dawson Reference Dawson1994). There is reason to be skeptical of the extent to which Black Americans view non-American Blacks as co-racial, especially when discussing issues unrelated to racism and racial discrimination (Smith Reference Smith2014), and linked fate is often used to evaluate whether or not American Blacks view Black immigrants and refugees as co-racial (Bunyasi and Smith Reference Bunyasi and Smith2019). We include this as a control variable to account for variation in the degree to which respondents view refugees as co-racial.

To measure the dependent variable, attitudes toward refugees, we include three conceptual dimensions of support: general attitudes toward refugees (including the belief that refugees should be welcomed to the United States, that they enrich U.S. culture, and that they should have the same rights as U.S. Citizens), action-based questions (how willing the respondent would be to take action to assist refugees in their resettlement, donating money, or calling representatives on their behalf), and economic threat questions (the belief that refugees take jobs from U.S. citizens, that refugees come here to take advantage of welfare programs, or that they are a drain on society). Each of these questions was asked on a scale of agreement on a 6-point Likert scale, from strongly agree to strongly disagree. Responses to these questions are coded such that a higher value indicates a higher level of support for refugees and are then compiled into an additive index (see Figure 4 for index distribution). ^{Footnote 9,Footnote 10}

Figure 4.

Distribution of support for refugees index.

To ease our ability to evaluate the relative strength of the substantive effects of these variables, the variables in these analyses were standardized. The independent variables are mean-centered at zero with each unit of change indicating one standard deviation. The dichotomous variables are coded with the categories as zero and one and are named to reflect the satisfied category of these variables.

Main Analysis

We conducted several statistical analyses, including parabolic t-tests, multiple ordinary least squares (OLS) regressions with interaction terms, and mediation analysis.Footnote ¹¹ For each analysis, we describe the results for white respondents and Black respondents separately. We begin by determining the overall effect of the racial priming treatment on support for refugees using a parabolic t-test (see Table 3). For white respondents, the mean support for refugees when primed with an image of a white refugee is 37.014, while support for refugees when primed with an image of a Black refugee is 36.161. While this does show slightly higher support for white refugees over Black refugees, the difference is not statistically significant (p = .149). For Black respondents, mean support for white refugees is 37.598, and mean support for Black refugees is 38.122. Again, while this shows a slight ingroup preference, it does not reach statistical significance (p = .283). The results of the t-test indicate that we do not find support for our first two hypotheses (H1A and H1B). Black respondents in this sample are, however, more supportive of refugees (regardless of the race primed) than white respondents (p < .01).These results are consistent with previous research showing that racial/ethnic minorities tend to be more supportive of other marginalized groups compared to whites (Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016; Sirin et al. Reference Sirin, Valentino and Villalobos2017).

Table 3.

T-test results

As previously discussed, group empathy is expected to have an independent effect on support for refugees that is robust to other controls including partisanship. To test this, we ran two multiple OLS regressions, one for white respondents and the other for Black respondents (see Figure 5 for coefficient plots).

Figure 5.

Coefficient plot for baseline models.

Note: Effects are shown per one standard deviation change. Dichotomous variables are shown as the difference between 0 and 1.

For white respondents, group empathy has a strong, positive, and statistically significant effect on support for refugees. For each standard deviation increase in group empathy, support for refugees increases by over 5 points (on an overall scale of 44 points). This effect is robust to all controls. Republican partisanship is associated with a significant reduction in support for refugees, although to a lesser substantive degree than group empathy. Age is negatively associated with support for refugees, while education level is positively associated with support for refugees. Both of these effects are statistically significant. Linked fate with other whites is also significantly associated with a decrease in support for refugees. The racial prime and gender are not significant. It is worth noting that group empathy has, by far, the largest standardized effect on support for refugees. Finally, the adjusted R² suggests that this model explains 65% of the variance in support for refugees for white respondents.

For Black respondents, we observe a similar empirical pattern. Group empathy has a very large, positive, and statistically significant effect on support for refugees. Republican partisanship has a significant, negative effect on support for refugees, but to a lesser degree than it did for white respondents. For white respondents, the per-standard deviation decrease in support was 3.758, while for Black respondents, it is 1.410. Age is negatively associated with support for refugees, while education is positively associated. Black women are more supportive of refugees compared to Black men.

Linked fate among Black respondents has the opposite relationship as it did for white respondents, where it significantly improves support for refugees. Black Americans who feel that the fate of all Black individuals within the country influences their own fate (suggesting a stronger connection to their racial ingroup) support refugees more than those with a weaker ingroup connection.Footnote ¹² One way to interpret this finding is that public opinion about refugees is about majority-minority group politics. Support for marginalized minority groups could hinge on both an individual’s position within hierarchical, societal structures and their beliefs about their linked fate in relation to other majority or minority groups.

All of the explanatory variables in the OLS regression for Black respondents are statistically significant with the exception of the racial prime. For the Black sample, the adjusted R² suggests that 42% of the variation in support for refugees is explained by our model. While our models for each racial sample explain a significant amount of variation in our dependent variable, the model fit is better for white respondents (65%) than it is for Black respondents (42%).

To determine the extent to which the effect of group empathy is conditional on the racial prime, we conducted multiple OLS regressions with interactions. Beginning with white respondents, the interaction between group empathy and racial priming is statistically significant (p < .05). While group empathy overall is clearly meaningful, its effect is moderated by racial priming as hypothesized: for those with low group empathy, seeing a racially incongruent prime caused a significant reduction in support for refugees, as compared to those who saw a racially congruent prime (see predicted values in Figure 6). For those with higher group empathy, the racial prime had no effect.

Figure 6.

Predicted values of support for refugees, white sample.

We observe a different relationship among Black respondents. As shown in Figure 7, the effect of group empathy on support for refugees is not conditional on the primed race. Group empathy is a strong, positive influence on support for refugees regardless of whether the refugee is racially congruent or incongruent. In sum, while there is a statistically significant ingroup racial bias among white respondents with low empathy, this does not hold true for Black respondents whose support for refugees does not meaningfully change based on the race linked with refugees. This further suggests that “racialized empathy” is a more important explanatory factor for white Americans than for Black Americans.

Figure 7.

Predicted values of support for refugees, black sample.

Our analysis does not allow us to explicitly test the mechanisms that explain why Black respondents are unaffected by the racial prime, however, we point to two pieces of evidence in the existing literature to make sense of this finding. First, Black Americans’ baseline support for immigrants and refugees is high relative to white Americans (Kinder and Sanders Reference Kinder and Sanders1996; Brader et al. Reference Brader, Valentino, Jardina and Ryan2010). Existing work on empathetic responses to vulnerable immigrants shows that while Black Americans are generally a little more supportive of racial minority immigrants, they still show support for white immigrants in distressing situations (Sirin, Valentino and Villalobos Reference Sirin, Valentino and Villalobos2016; Sirin, Villalobos and Valentino Reference Sirin, Villalobos and Valentino2016). Our finding aligns with work that argues that Black and white Americans bring different lenses to issues of immigration. The race of immigrant-origin people is not as salient to Black opinion formation as it is for whites (Carter Reference Carter2019). This is reflected in our data, with Black respondents exhibiting more support for refugees compared to white respondents regardless of the race of the refugee they are primed to consider.Footnote ¹³ Second, Black Americans may not feel a strong enough racial ingroup attachment to Black refugees compared to white refugees because they do not necessarily view African migrants as co-racial (Waters Reference Waters2022; Unnever and Gabbidon Reference Unnever and Gabbidon2015). This characteristic could dampen the ingroup effect that increases support for Black refugees above a generally supportive baseline.

Mediation Analysis

While the positive and direct effect of group empathy on support for refugees stands out across model specifications, partisanship also has a clear effect on support for refugees (albeit a stronger predictor of white attitudes). To test whether partisanship explains variation in empathy for refugees (H3), we conduct a mediation analysis where partisan identity is the treatment, group empathy for refugees is the mediator, and support for refugees is the outcome. The traditional framework for causal mediation analysis requires that three conditions are met: 1) The treatment has a statistically significant effect on the outcome in the expected direction, 2) The treatment has a statistically significant effect on the mediator in the expected direction, and 3) When the mediator is included in a model with the treatment and outcome, the effect of the treatment on the outcome is substantially reduced (Baron and Kenny Reference Baron and Kenny1986). We use Kosuke Imai and colleagues’ strategy for mediation analysis (Imai et al. Reference Imai, Keele, Tingley and Yamamoto2010, Reference Imai, Keele, Tingley and Yamamoto2011).

The results reported in Table 4 suggest that while the effect of partisanship on support for refugees is partially mediated by group empathy, especially for white respondents, this effect is relatively weak compared to the direct effect of partisanship. For both Black and white respondents, the more Republican they are, the less supportive they are of refugees (see columns 1 and 4). These results satisfy the first condition of mediation—the treatment has a statistically significant association with the outcome in the expected direction. We also find support for the second mediation requirement that the treatment (partisanship) has a statistically significant effect on the mediator (group empathy). As shown in columns 2 and 5 in Table 4, increases in the Republican partisanship scale lead to decreases in group empathy for refugees. We find partial support for the third causal mediation condition. When the mediator (group empathy) is introduced to the basic treatment model where partisanship predicts support for refugees, the effect of partisanship decreases (see columns 3 and 6). However, the reduction is minimal for both the white and Black samples, and the effect of partisanship on support for refugees does not reduce in statistical significance.

Table 4.

Mediation analysis: group empathy partially mediates effect of partisanship on support for refugees

Note: *p $ \lt $ .1; **p $ \lt $ .05; ***p $ \lt $ .01; OLS regression.

Next, we turn to the formal mediation analysis. If partisanship affects attitudes toward refugees through group empathy, then the Average Causal Mediation Effect (ACME) would be non-zero and statistically significant and the Average Direct Effect (ADE) would be near zero and either not statistically significant or nearly null. As displayed in Figures 8 and 9, there is some evidence that group empathy for refugees mediates the effect of partisanship on support for refugees. The ACME estimate of the indirect effect is non-zero and statistically significant for both Black and white respondents. The size of the indirect effect estimate is larger for the white sample suggesting group empathy mediates more of the partisanship attitudes toward refugees association compared to the Black sample.

Figure 8.

Mediation analysis for Black respondents.

Figure 9.

Mediation analysis for white respondents.

However, the direct effect of partisanship on support for refugees is much larger compared to its indirect effect through group empathy for both racial groups. This suggests that while group empathy could be one of the mechanisms through which partisanship shapes public opinion about refugees, it is not a primary causal mechanism. Further experimental research is required to tease out the roles of partisanship, group empathy, and attitudes toward refugees.