CONCEPTUALIZING RACIAL CONTEXTS

Racial group membership has sweeping political implications in the United States, shaping everything from turnout patterns to policy outcomes (Hutchings and Valentino Reference Hutchings and Valentino2004). But racial groups themselves are far from objective phenomena. Rather, race and the boundaries between groups are informed by social, political, historical, and economic dynamics (Anoll Reference Anoll2022; Omi and Winant Reference Omi and Winant2014). Noting this, scholars have worked to identify the various factors that contribute to the creation and definition of race in America (e.g., Davenport Reference Davenport2018).

One such contributor is the geographic arrangement of peoples. Racial segregation in the United States is deeply rooted in systemic factors that span centuries. White Americans have played a key role in creating and sustaining geographic segregation through a combination of historical actions, policies, and social practices (Lopez Reference Lopez1996; Ray Reference Ray2019). The establishment of slavery in the colonial period laid the foundation for racial divisions, with whites holding a position of power and privilege. Even after the abolition of slavery, the ideology of white supremacy persisted, leading to discriminatory practices and policies, such as Jim Crow laws that enforced racial segregation in public facilities, schools, and housing (Katznelson Reference Katznelson2005; Massey and Denton Reference Massey and Denton1993).

That this geographic arrangement of peoples helps explain politics was established early in political behavior research. Key (Reference Key1949) showed that as the proportion of the white in-group diminished in southern counties and states, negative out-group attitudes and turnout among whites increased. Many since have examined how the composition of geographies ranging from states to census blocks affects political action (e.g., Fraga Reference Fraga2018; Trounstine Reference Trounstine2018), arguing that increasing diversity in the United States is driving right-wing populism, returns to old-fashioned racism, and white racial animus (Craig and Richeson Reference Craig and Richeson2014; Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2018). That is, increased contextual diversity leads to increased threat.

Writing during the same period, Gordon Allport came to a very different conclusion. “People in the southern states may think they know the Negro,” he wrote, “But their contacts are casual, or else firmly frozen into superordinate-subordinate relationships.… In contrast to casual contacts, most studies show that true acquaintances lessen prejudice” (Allport Reference Allport1955, 263–4). Focusing on these close contacts, scholars since have shown that cross-race relationships can lead to changes in political behavior—often for the better. Cross-cutting relationships can reduce prejudice, leading to support for policies that aid out-groups (Walker, McCabe, and Matos Reference Walker, McCabe and Matos2022).

The juxtaposition between Key and Allport is often characterized as opposition: one tradition suggests out-group contact increases threat, while the other says it lowers prejudice. But the authors—and the traditions that followed from them—are not so much in disagreement as they are simply examining different levels of context. Key and his descendants focus on geographic racial context, or the physical arrangement of racial groups within a defined geographic space including neighborhoods, cities, states, and even nations. Those in the Allport tradition consider social racial context, or how the composition of (close) social ties and intimate contact with the out-group shape people’s lives.

As these scholarships have grown, the meaning of context has grown with them—and amid this burgeoning, the distinction between concepts and the relationship across contexts has become jumbled. In their work on contextual measurement, Wong et al. (Reference Wong, Bowers, Rubenson, Fredrickson and Rundlett2020, 161) note that “scholars tend to conceptualize ‘community’ as place, space, and networks of social ties and allegiances.” This conceptualization conflates geographic context (“place, space”), social context (“social ties”), and psychological context (“allegiances”). Indeed, scholars now use a variety of measures—for instance, network analyses, census data at myriad geographic levels, and self-reported neighborhood boundaries—to make claims about how context generally shapes political attitudes. But the choice of measure is often based on whatever contextual variable is most readily available with little justification for this choice or engagement with alternatives (Wong et al. Reference Wong, Bowers, Williams and Simmons2012). There remains little empirical engagement with the relationship between types of racial embeddedness more generally in the United States, despite scholars acknowledging that the political effects of geographic racial context may be due to “a complicated set of overlapping contexts” (Baybeck Reference Baybeck2006, 387).

So, what is the relationship between racial context types in the United States? One possibility is that contexts are tightly entwined, with one funneling into and creating through its structure the next. If this is the case, then distinguishing empirically between contexts is of little consequence. Rather, any given measure can proxy for the others to capture generalized in-group contact. In this vein, Lawler, Ridgeway, and Markovsky (Reference Lawler, Ridgeway and Markovsky1993, 272) argue that geographic spaces like neighborhoods tend to “generate the reasons for and conditions of encounters, [and] determine who encounters whom.” Geographic context contains schools, businesses, and places of worship that bring people into contact with each other, allowing encounters to occur and repeat over time (see also Eulau and Rothenberg Reference Eulau and Rothenberg1986; Huckfeldt Reference Huckfeldt1983). This repetition may produce stable social networks: people find spouses, form friendships, and develop groups of others to rely on. In this theory, one level of context (geographic) directly creates a second type of context (social). If individuals live primarily around others who share their race, their geographic context may produce social networks that similarly reflect this homogeneity.

The racial composition of geographic and social contexts may further contribute to the formation of a third kind of in-group embeddedness: psychological racial context, or the degree to which individuals feel embedded in their racial group (McClain et al. Reference McClain, Johnson Carew, Walton and Smith2009; Wong Reference Wong2010). From past works, we know that in-group mental primacy can vary across individuals, with the strength of racial-group closeness a central predictor of vote choice, policy attitudes, and turnout (Conover Reference Conover1984; Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986; Wong Reference Wong2010). Further, this in-group mental primacy is shaped by both external constraints and internal choices; it is contextual in its formation and expression (Pérez Reference Pérez2021). When an individual has a stronger in-group identity and sense of closeness, they center their group in their decision-making and are more responsive to in-group cues and needs.

Psychological racial context in our conception, then, captures how proximate and accessible racial groups are in the mind as individuals make decisions—much like geographic context captures how proximate and accessible racial groups are in physical space. We define racial context as the setting in which people of different ethnoracial groups form their inter- and intragroup attitudesFootnote ¹—that is, racial context is a frame that surrounds experience and “provides resources for its appropriate interpretation” (Duranti and Goodwin Reference Duranti and Goodwin1992, 3). By this definition, the degree to which individuals invoke the group in their mind as they navigate the world and use it as a frame for interpreting events captures a unique racial context that is distinct from the geographic dispersion of peoples or close social ties.

The racial composition of one’s geographic and social context may contribute to the strength of this in-group racial identity, or psychological racial group context. Enos (Reference Enos2017, 4) argues that “geography penetrates our psychology—it affects the very way we perceive other groups—with these changes in perception, it affects our behavior.” Lawler, Ridgeway, and Markovsky (Reference Lawler, Ridgeway and Markovsky1993) contend that, “areas of high network density foster frequent interaction and give members the opportunity to compare their attitudes to those of others and to make adjustments” (274). That is, repeated interactions through close social ties may begin to shape the internal view of self. In teaching people about group norms, changing reference prototypes, and shaping the perceived differences between groups, identities emerge and are given strength and meaning (Anoll Reference Anoll2022). The racial composition of one’s geographic and social context then may alter one’s psychological racial context via the salience of one’s in-group racial identity. If this is the case, psychological racial context may be a downstream mechanism through which geographic and social racial contexts function.

In this theory, racial contexts in the United States are nested around each other: geographic, social, and psychological racial contexts funnel into and create each other. If this is the case, the three measures would be highly correlated. We visualize this possibility on the left side of Figure 1. Here, the composition of geographic space funnels into the composition of close social ties, which in turn informs the strength of psychological racial context. Considering work on the hyper-segregation of race in the United States (Massey and Denton Reference Massey and Denton1993), we might expect that most Americans, regardless of racial group membership, will be highly embedded in their in-group context across levels.

Figure 1. Possible Variations in the Degree of Nestedness in Racial Contexts

Alternatively, levels of in-group racial embeddedness for individuals may vary by contextual type. Rather than geographic, social, and psychological contexts nesting around each other, unique factors may shape each so that high in-group embeddedness in one context does not translate to the same embeddedness in others. This possibility may be facilitated by elements of the modern world. Innovations in technology allow social ties to flourish regardless of geographic proximity, leading some to note that the internet and social media platforms have led to the “death of distance” (Scellato et al. Reference Scellato, Noulas, Lambiotte and Mascolo2011). Although geographic proximity brings people into a shared physical space, this does not necessarily produce meaningful interaction (Michener Reference Michener2013). Further, a feeling of psychological connection to a group may develop from factors separate from immediate and current social or geographic space. Influences such as early socializing experiences (Jennings and Niemi Reference Jennings and Niemi1968), interactions in the work place (Mutz and Mondak Reference Mutz and Mondak2006), and media sources (Dávila Reference Dávila2012) may collectively contribute to feelings of group attachment in ways that conflict with social or geographic embeddedness. Additionally, living in a geographic context composed of primarily out-group members, or maintaining diverse social ties, may make one’s own racial identity more salient—increasing a sense of psychological co-racial connection that conflicts with a relatively diverse social and geographic environment.

If this is the case, the racial composition of place, social ties, and in-group allegiances may not be closely related, but instead introduce distinct or even conflicting pressures. Rather than proxying for each other, individuals could experience high embeddedness in their racial group in one context, which may send signals about group norms and political attitudes, that conflict with pressures from other contexts where out-groups are more salient. We visualize this possibility of non-nested racial contexts on the right side of Figure 1, where no individual experiences high levels of embeddedness in geographic, social, and psychological contexts simultaneously. In this scenario, each contextual dimension may produce independent effects on out-group opinions or create interactions across contextual arrangements.

Between fully nested and non-nested theories of group embeddedness is a middle ground, which we display with Figure 1’s center image. In this heterogeneous distribution, some individuals may experience high levels of concentric and complementary contextual factors; such individuals would live in geographic spaces primarily comprised of in-group members, maintain close social ties with the in-group, and feel close to other co-racial members. But other individuals may experience cross-cutting cleavages, where despite being highly embedded in one or two contexts, they have low embeddedness in another.

Previous work has sought to measure geographic, social, and psychological racial contexts in the United States and its effect on political outcomes (Barth, Overby, and Huffmon Reference Barth, Marvin Overby and Huffmon2009; Barreto and Pedraza Reference Barreto and Pedraza2009). However, few have directly compared political effects across different racial contexts. If racial contexts are indeed highly nested, then understanding the political effects of one level of racial context sheds much light on the likely political impact of the other two levels. If, however, racial contexts are only partially nested or completely non-nested, understanding the political consequences of each contextual type becomes more complicated. Specific ethnoracial groups may further vary in how nested or non-nested their geographic, social, and psychological racial contexts are. Ethnoracial groups experience hugely different constraints and opportunities in their ability to move, associate with one another, and “forget”/“remember” their group (Jardina Reference Jardina2019; Wilson Reference Wilson2011). The relationship between contextual measures, then, should be considered uniquely across U.S. ethnoracial groups.

POTENTIAL VARIATION ACROSS ETHNORACIAL GROUPS

Different racial groups in the United States vary in their relative size and relationship to power, which may produce variation in their degree of contextual nestedness. Take, for instance, the contextual arrangement of white Americans and their position in the racial hierarchy. White Americans, as a majority population, have often wielded political and economic influence, implementing Jim Crow laws and other discriminatory policies such as redlining, which designated areas with predominantly Black and Latino populations as high-risk for mortgage lending. This systematically disadvantaged Black and Latino communities in terms of housing and financial opportunities (Lopez Reference Lopez1996). The legacy of these actions continues to shape patterns of geographic segregation in cities across the United States, contributing to disparities in access to resources, educational opportunities, and economic mobility among different racial groups (Chetty et al. Reference Chetty, Hendren, Jones and Porter2020).

As a result, the process through which white Americans have become embedded within psychological, geographic, and social racial contexts is fundamentally different than that of racial minorities. A long history of de jure and de facto discrimination means white people tend to live in geographic areas that are overwhelmingly white (Logan and Stults Reference Logan and Stults2011) and marry and build families with others who are also white (Davenport Reference Davenport2018). Yet whites’ high social status and lower level of exposure to other racial groups have historically made whites less likely to develop a strong sense of group solidarity (Sears and Savalei Reference Sears and Savalei2006). However, scholarship has shown a shift in recent years, as white identity has risen in response to changing racial demographics (Craig and Richeson Reference Craig and Richeson2014). As Jardina (Reference Jardina2019) argues, white identity has become influential in shaping individuals’ political views on issues such as immigration, affirmative action, and diversity.Footnote ² Given this, white Americans may, on average, experience high levels of embeddedness in their geographic and social racial contexts—with most living in majority-white neighborhoods and maintaining close social ties with mostly other white people—but exhibit moderate levels of embeddedness in their psychological racial context.

In contrast, geographic, social, and psychological racial contexts for Black Americans may be both very nested and connected to key political orientations. Importantly, per Dawson (Reference Dawson1994), Black Americans evince high levels of co-racial psychological embeddedness via their belief in racial linked fate—the collective sense of interconnectedness and shared destiny among individuals within the Black community—which in turn deeply shapes Blacks’ political behavior. White and Laird (Reference White and Laird2020) argue that high levels of segregation among Black Americans in both social and geographic space contribute to “racialized social constraint”: a process of norm enforcement that involves compliance with group-based political behavior, including political unity and the pursuance of shared interests.Footnote ³

Like Black Americans, Latino Americans have endured a history of legal segregation. In addition to being subjected to discriminatory housing practices, workplace discrimination has limited many Latinos’ economic opportunities, contributing to the concentration of Latinos in certain low-income neighborhoods. Yet, given the continual flow of migrants from Latin America into the United States, the forces steering geographic segregation among Latinos differ from those of Blacks in important ways. Notably, immigration policies throughout U.S. history have had a profound impact on the settlement patterns of Latino populations. For example, border enforcement policies aimed at curbing unauthorized immigration have influenced Mexican migrants’ border crossing routes and destination choices (Bohn and Pugatch Reference Bohn and Pugatch2015; Massey, Durand, and Malone Reference Massey, Durand and Malone2002). Immigrants may also opt to settle in neighborhoods with established Latino communities for better access to cultural resources (such as ethnic grocery stores or businesses that cater to Latino interests) and to live around people who share a similar language and cultural background (Kershaw, Albrecht, and Carnethon Reference Kershaw, Albrecht and Carnethon2013), increasing in-group social and geographic embeddedness. In turn, research suggests that the political effects of geographic, social, and psychological racial contexts are intertwined for Latinos: residing in a majority-Latino district can empower Latino voters and boost Latino turnout (Barreto, Segura, and Woods Reference Barreto, Segura and Woods2004) and the effects of certain voter mobilization strategies have also been shown to be conditional on the strength of group identity and community-level resources (Valenzuela and Michelson Reference Valenzuela and Michelson2016).

U.S. immigration policies have also importantly shaped the distribution of Asian communities across the United States over time. The Chinese Exclusion Act (1882) restricted immigration and impacted the settlement patterns of Chinese immigrants. Following the Vietnam War, refugee policies shaped resettlement patterns of Southeast Asian communities, with many settling in specific areas where refugee assistance programs were established, such as California and Texas. More recent economic immigration policies favoring skilled immigrants have contributed to the settlement of Asian immigrants, particularly from India and China, in regions with tech industries and research centers. And political science research shows that Asian Americans’ geographic context and immediate social networks uniquely shape their partisan preferences (Raychaudhuri Reference Raychaudhuri2018). Neighborhood contextual factors, such as the presence of co-ethnic immigrants, pose important consequences for Asian Americans’ voter turnout (Tam Cho, Gimpel, and Dyck Reference Tam Cho, Gimpel and Dyck2006), and a key predictor of vote choice for Asian Americans is county-level partisan context—the effects of which can be heightened by social integration into local communities (Raychaudhuri Reference Raychaudhuri2020).

Compared to Black Americans, both Asians and Latinos are, on average, more residentially and occupationally integrated with whites and more likely to intermarry with whites (Lee and Bean Reference Lee and Bean2010), suggesting lower levels of co-racial embeddedness in geographic and social contexts compared to Blacks. However, due to high rates of recent migration and xenophobic stereotypes, both Asian Americans and Latino Americans commonly experience racial discrimination (Kuo, Malhotra, and Mo Reference Kuo, Malhotra and Mo2017; Zou and Cheryan Reference Zou and Cheryan2017), which may produce high levels of psychological in-group closeness.

To summarize: exclusionary legal and social practices, including racially discriminatory housing policies and immigration laws, have importantly shaped the in-group embeddedness of Black, Latino, and Asian Americans. Considering variation in U.S. ethnoracial groups’ historical experiences, positionality, and group size, we expect variation in the nestedness of geographic, social, and psychological contexts across groups. Specifically, for white Americans, we expect a high level of embeddedness in geographic and social contexts, and moderate level of embeddedness in psychological context. For Asian Americans, we expect a moderate level of embeddedness in geographic and social contexts, and high embeddedness in psychological context. For Black and Latino Americans, we expect high embeddedness on all three contextual measures.

Such ethnoracial variation would have implications for how we measure, conceptualize, and test the influence of racial context on politics. If contexts are neatly nested for the vast majority of Americans, regardless of racial group membership, then one measure of racial context can reasonably stand-in for the next. But if, as we hypothesize, racial embeddedness varies across contextual levels and is cross-cutting for some segment of the U.S. population, then contextual scholars will need multiple measures to confirm or deny a relationship between racial in-group embeddedness and political outcomes. Further, if nestedness varies by racial group, then contextual effects may also differ depending on group membership.

MEASURING RACIAL CONTEXT(S)

To examine racial contexts, we use the Participatory Social Norms Survey (PSNS) (Anoll Reference Anoll2022). Conducted in March 2018 in both English and Spanish through the online platform GfK, the PSNS boasts a nationally representative sample of each of the four largest U.S. ethnoracial groups: whites, Latinos, Blacks, and Asians.Footnote ⁴ In total, our study includes 1,000 Black Americans, 1,000 white Americans, 996 Latino Americans, and 695 Asian Americans.Footnote ⁵ The PSNS is well suited for our purposes because it includes large samples of four racial groups and multiple measures of racial context.

We specify geographic racial context as the proportion of each respondent’s residential census tract that is co-racial. Scholars within the geographic context tradition have noted that relationships can change depending on the geographic level of aggregation used, a problem referred to as the “modifiable areal unit problem” (MAUP). Like we consider the relationship between different contexts, scholars have attempted to understand the relationship between different areal units, including state, county, census tract, or zip code (Fraga Reference Fraga2016; Hopkins Reference Hopkins2010).

We select census tract because it is a relatively small geographic unit that follows legal boundaries. Scholars have suggested that racial contact is likely to be most visible at smaller geographic units and that localized geographic measures help more precisely identify relationships (Hersh and Nall Reference Hersh and Nall2016; Oliver and Wong Reference Oliver and Wong2003). However, we are attentive to the MAUP, and as Tam Cho and Baer (Reference Tam Cho and Baer2011) recommend, we examine whether our estimates are sensitive to different geographic levels including zip code, which others argue is the areal unit that most closely represents citizens’ concept of “their community” (Velez and Wong Reference Velez and Wong2017). We find similar results when we measure geographic context at the zip code level (see Supplementary Figure A-6).

To calculate geographic racial context at the census tract level, census tract indicators for PSNS respondents were merged with data from the U.S. Census Bureau’s 2013–2017 American Community Survey 5-Year Estimates, which provide the racial composition and population size of respondents’ census tracts (Manson et al. Reference Manson, Schroeder, Van Riper, Kugler and Ruggles2022). From these data, a single indicator was created to represent the proportion of a respondent’s census tract that was single-race Asian, Black, white, or Latino, matched to the respondent’s self-reported race.

Social racial context refers to the degree to which an individual maintains intimate social ties that are comprised primarily of racial in-group members. These ties tend to be close, frequently interacted with individuals that are socio-emotional, rather than purely instrumental, in nature (Marsden Reference Marsden1987; McPherson, Smith-Lovin, and Brashears Reference McPherson, Smith-Lovin and Brashears2006). To empirically specify this concept, we use a name-generator question focused on egocentric networks. Respondents were asked, “From time to time, most people discuss important matters with others. Looking back over the last six months, who are the people with whom you discussed matters important to you?” Respondents were provided space to generate up to five first names or initials of individuals who fit this description.Footnote ⁶ On the next page, respondents were asked,“What is the race/ethnicity of each person?” with the initials or names from the previous question inserted. Respondents were provided with a check-box matrix of racial groups to select for each reported contact. From these data, we calculate the proportion of each respondent’s reported network that is co-racial. We do this by dividing the number of connections the respondent reports as sharing their race from the total number of close connections reported (up to 5), producing a measure that ranges from 0 to 1.

For psychological racial context, we assess how central a group is to an individual’s thought process by relying on a measure of self-reported group closeness (Leach et al. Reference Leach, van Zomeren, Zebel, Vliek, Pennekamp, Doosje and Ouwerkerk2008; Wong Reference Wong2010). This measure corresponds to our conceptualization of psychological racial context: we contend that individuals who feel closer to other members of their racial group will be more likely to think of their racial group, as well as its norms and considerations, as they navigate through the world. Respondents reported how close they feel to each of the four largest ethnoracial groups, including their own. Four response options ranged from “not at all” to “very close.” From this question, we create an indicator of co-racial group closeness for each respondent and rescale the measure from 0 to 1, where 0 indicates an answer of “not at all close” and 1 indicates an answer of “very close.”

Some readers may wonder whether self-reported beliefs about the racial composition of one’s community (e.g.,Velez and Wong Reference Velez and Wong2017; Wong et al. Reference Wong, Bowers, Williams and Simmons2012) better capture psychological racial context. Such a measure, we argue, assesses subjective geographic racial context—that is, the dispersion of peoples in physical space—and not psychological racial context—that is, the centrality of the group in the mind. Our goal in drawing closeness into our model is to connect three disparate literatures—one on the geographic distribution of peoples, a second on social networks, and a third on group closeness—under a single theoretical frame to identify whether and how these concepts relate to one another.

Figure 2 shows the distribution of these three contextual measures for each group. The x-axis provides the measure-specific categories of the distribution; the y-axis plots within-group proportions. The plots show that racial contextual embeddedness varies significantly both by contextual level and by racial group. For white Americans, the distribution of geographic racial context is left-skewed, with 59% of this group living in a geographic context that is 80%–100% white. In sharp contrast, the distribution is right-skewed for Asian Americans, 68% of whom live in census tracts that are less than 20% Asian. Blacks and Latinos are in between: 36% of Black Americans live in census tracts that are less than 20% Black; and 34% of Latinos live in census tracts that are less than 20% Latino.

Figure 2. Distribution of In-Group Embeddedness across Racial Contexts

Note: Not all respondents reported five close ties; thus, x-axis categories reported in panel (b) also represent off-category groups: $ 0=0 $ ; $ 0.2=0.2,0.25 $ , and $ 0.33;0.4=0.4 $ and $ 0.5;0.6=0.6 $ and $ 0.67;0.8=0.75 $ and $ 0.8;1=1 $ . Full results and sample sizes can be found in Supplementary Table A-2.

Despite this wide variation in the modal categories of each distribution, mean evaluations demonstrate the continued legacy of geographic racial segregation in the United States, which has often resulted from purposeful policy choices made by white elites. While white Americans make up 61% of the population, they live in spaces that are on average 77% white; Black Americans and Latinos comprise 13% and 18% of the population in the United States, respectively, but live in spaces that are roughly 40% co-racial;Footnote ⁷ and Asian Americans, who make up 6% of the population, live in spaces that are on average 17% Asian.

The distribution of racial group embeddedness at the level of social racial context again shows the stark realities of contemporary segregation, as Figure 2b demonstrates. Examining co-racial social embeddedness, we find that the modal category for each group on a 0–1 scale is 1, signifying an entirely racially congruent network of intimate ties. More specifically, 81% of white Americans, 62% of Black Americans, 44% of Asian Americans, and 42% of Latino Americans report that their close social networks are entirely co-racial. Still, variation exists across racial groups in the degree of average co-racial social embeddedness. White Americans have, by far, the most homogeneous social ties, with an average of 91% co-racial close ties. This estimate outpaces the next most socially segregated group, Black Americans, by nearly 14 percentage points ( $ \overline{x} $ Black = 0.77, $ \overline{x} $ Latino = 0.62, $ \overline{x} $ Asian = 0.60).

The distributions for psychological racial context (Figure 2c), too, show variation across groups and differences when compared to the other two contextual measures. White, Black, and Latino respondents show roughly similar levels of average psychological in-group embeddedness, with statistically indistinguishable means ranging from 84% to 86% ( $ p>0.05 $ ). Asian Americans demonstrate weaker psychological attachments to the racial in-group, with a statistically distinct average of 77% ( $ p<0.05 $ ). Still, all four distributions are on average left-skewed, suggesting that most Americans, regardless of their racial group membership, feel “very close” to in-group members.

The distributions of these three measures largely support the hypotheses we advance about how embeddedness in racial contexts may vary across groups. White Americans are in fact highly embedded in their ethnoracial group at the geographic and social levels. Black and Latino Americans also exhibit high levels of co-racial geographic and social embeddedness—especially compared to their proportion in the population—and are highly embedded in their racial group psychologically. Asian Americans, in contrast, are somewhat less likely to live in geographic environments that are predominantly Asian, but they do exhibit relatively high levels of psychological co-racial embeddedness.

One notable divergence from our hypotheses concerns white Americans’ psychological racial context: Figure 2c shows white Americans have a high level of psychological in-group embeddedness, rivaling the level we see among Black and Latino Americans. That white Americans are so connected to their racial in-group psychologically reflects more recent work on the strength of white racial identity (Craig and Richeson Reference Craig and Richeson2014; Jardina Reference Jardina2019).

What is the relationship between these contextual measures? We turn to this question next.

DEGREE OF NESTEDNESS IN AMERICANS’ RACIAL CONTEXTS

We have argued that extant literature provides competing hypotheses about how racial contexts—geographic, social, and psychological—overlap in the United States. In this section, we test the degree to which our three contextual measures nest around each other for Americans of different racial groups. We do so by examining the correlations between context measures and the percentage of each racial group categorized as “highly embedded” in multiple racial contexts.

We begin with a test of correlations, examining the strength of the relationship between each contextual pair. Large, positive coefficients suggest that racial contexts match the nested theoretical model, with each measure strongly related to the other. Small or negative coefficients suggest that racial contexts are not nested, or are even negatively related.

Table 1 provides Pearson’s correlation coefficients for each of the measurement pairs by racial group. Positive and significant coefficients on most of the estimates suggest that racial contexts are indeed related; however, the relatively small magnitude on all 12 point estimates provides evidence this relationship is weak. Taking the geographic–social correlation coefficients first, we see the point estimates range from 0.16 (white) to 0.39 (Latino), with Black (0.28) and Asian Americans (0.29) arrayed in the middle. All four estimates reach statistical distinction from zero ( $ p<0.05 $ ) but fall well below the conventional standard of 0.85 for testing the interrelatedness of measures (Campbell and Fiske Reference Campbell and Fiske1959).

Table 1. Correlation of Racial Embeddedness across Measures

Note: Entries represent Pearson’s correlation coefficients. Sample size for each test in parentheses. * $ p<0.05 $ .

Results from the social–psychological and geographic–psychological paired tests show even weaker relationships between these contextual measures. In-group embeddedness at the social contextual level is correlated with psychological embeddedness at a rate of 0.11 (white) to 0.24 (Latino), depending on the group. Geographic racial context and psychological racial context correlate at rates ranging from 0.03 (white) to 0.13 (Latino), with only estimates for Asians, Blacks, and Latinos reaching statistical distinction from zero ( $ p<0.05 $ ). For each of the groups, the relationship between social and psychological contexts is weaker than geographic and social contexts, and the relationship between geographic and psychological contexts is consistently the weakest across groups.

Still, some directionality of influence is apparent in the table. Generally, the relationships measured in the first two columns (geographic–social; social–psychological) are stronger than those found in the third column (geographic–psychological). This suggests a possible mediating effect of social ties between geography and psychological embeddedness: geographic context may funnel people into social contexts, which in turn shape psychological connectedness to groups. But the weak coefficients suggest that if funneling is happening, there is leaking and are sources of dilution along the way.

Collectively, Table 1 provides evidence that levels of in-group embeddedness in geographic, social, and psychological racial contexts are only weakly related to each other. Racial contexts appear to more closely approximate non-nested and cross-cutting cleavages than neatly nested circles that proxy for each other. Still, the weak positive relationships suggest there are likely at least some Americans who experience high levels of embeddedness across contexts.

We turn next to estimating this relationship. We identify the percentage of each group who experience high in-group racial embeddedness across each contextual level. An individual is coded as “high” separately for each context type if they are within the weighted top 25% for their ethnoracial group.Footnote ⁸ Figure 3 shows the percentage of each racial group who experience high levels of in-group embeddedness across various contextual types. We visualize the findings with weighted venn diagrams that show the percentage of each racial group experiencing nested versus non-nested contextual embeddedness (see also Supplementary Table A-3).

Figure 3. Overlap between Racial Embeddedness Measures

Note: Respondents are considered “high” in a given embeddedness category if they are in the weighted top quartile for their race. Respondents with missing values are excluded. All measures include sample weights. Full results and sample sizes can be found in Supplementary Table A-3.

The figure demonstrates that between 6.8% and 16.3% of Asian, Black, Latino, and white Americans experience contextual in-group nestedness. That is, they live in geographic contexts comprised of primarily in-group members, their social ties are largely the same race as themselves, and they report high levels of psychological closeness to their racial in-group. The estimates also show that most Americans, regardless of race, experience some degree of variation in their contextual experience: 33% of white Americans, 35.8% of Black Americans, 37.4% of Latino Americans, and 43% of Asian Americans are highly embedded in their group on only one contextual level. For all non-white racial groups, psychological embeddedness is the most common; for white Americans, it is social embeddedness. An additional 26.1% of Asian Americans, 28.6% of Latino Americans, 36.2% of Black Americans, and 45.7% of white Americans are highly embedded in their group on two of the contextual measures.

The lack of complete overlap between contextual measures visualized in Figure 3 suggests that rather than being neatly nested contextual proxies, measures of context are often cross-cutting, varying in degree. There is some variation in nestedness across different racial groups; for example, white Americans are more likely to simultaneously experience high levels of embeddedness in their social and psychological racial contexts than are their non-white counterparts. However, we contend that the most notable finding in Figure 3 is how few Americans in all four racial groups fall into the center category of the venn diagrams. Most Americans, regardless of race, are weakly embedded in their racial group on at least one contextual level but strongly embedded on at least one other. Variation in the nestedness of racial embeddedness measures has implications for our understanding of their relationship to political outcomes. If geographic, social, and psychological racial contexts were in fact largely nested for most Americans, any measure could serve as a proxy for the others. But in the world we actually observe, racial contexts are largely cross-cutting. As a result, making inferences about the political effects of one level of racial context based on the effects of a different level leads to incomplete—or even inaccurate—understandings of how and when racial context matters for politics. Considering this, in the next section, we (re-)examine how co-racial contextual embeddedness uniquely relates to key political attitudes and policy preferences.

RACIAL CONTEXTS AND POLITICAL ATTITUDES

We assess the relationship between geographic, social, and psychological contexts and six political outcome variables, each selected because of their centrality in understanding political behavior and previous connections to context. We cannot establish the causal effect of racial context on behavior with our observational data, as the racial contexts in which people choose to live may be endogenous to their politics. However, the relationships can tell us whether contextual measures can proxy for each other or whether each is uniquely related to core political attitudes.

Our first two outcomes focus on orientations toward politics. First, we examine the relationship between racial contexts and external political efficacy, or the degree to which individuals think public officials care about people like them (Soss Reference Soss2000). Second, we consider partisanship, measured on a scale ranging from strong Republican to strong Democrat.Footnote ⁹ These attitudinal variables have been repeatedly shown to shape individuals’ orientations toward political participation, policy positions, and vote choice, making them central to understanding the American political landscape.

We also examine the relationship between racial context and four racialized policy positions. Namely, we consider attitudes about whether the federal government should increase the deportation of immigrants who are in the United States illegally (immigration); whether it is better if English is the only language used in public schools (English language); whether the police use the right amount of force in a given situation (policing); and whether the government should increase money spent on aid to the poor (aid to the poor). We measure both attitudes about undocumented immigrants and attitudes about English language in schools because they tap into different dimensions of immigration and immigrant attitudes: the former tends to reveal attitudes about Latino immigrants in particular, while the latter corresponds to attitudes about a wider swath of immigrants, including Asian immigrants. Scholars have suggested that because of the racialized nature of these policies, racial context proves especially important for explaining individuals’ positions on these topics—though, this varies by group (e.g., Carter Reference Carter2019; Gilens Reference Gilens2009; Wong Reference Wong2010). We test here how each unique measure of racial context may relate to positions on these policies across racial groups.

We predict variation in each outcome variable with a linear regression model that inserts one of the racial context variables—geographic, social, or psychological—as the primary independent variable.Footnote ¹⁰ We include in these regressions five individual-level covariates (age, gender, household income, educational attainment, and nativity/generational status) and three census tract-level variables (median income, proportion with at least a high school degree, and proportion foreign born).Footnote ¹¹ In line with best practices (Masuoka and Junn Reference Masuoka and Junn2013), and since contexts may relate differently to outcome variables across races, we run separate models for each racial group.

Our goal is to explore patterns that may exist between different contextual variables and politics across racial groups. We are not engaged in a hypothesis testing mission where each unique regression is taken as a stand-alone finding. Instead, we focus on whether some contextual variables are regularly more predictive of political outcomes than others and whether this variation occurs across racial groups. To visually highlight these patterns rather than specific point estimates, we present our results in Figure 4 as a heat map. In this plot, the y-axis identifies the race of individuals included in the model; the x-axis identifies the outcome variable for the model; and the panels identify the main independent variable: namely, geographic, social, and psychological embeddedness in a co-racial context. Pulling coefficients from each model, we visually represent the relative size and direction of a full-scale change in the independent variable (context) on the dependent variable (political outcomes). Blue tiles represent positive coefficients, while green points with a black border represent negative coefficients. Larger tiles represent relatively larger point estimates, with tile size scaled relative to the largest coefficient in any of the models run. Opacity reflects statistical significance: opaque tiles are estimates that were found to be significant (at $ p<0.05 $ ).Footnote ¹²

Figure 4. Effect of Racial Context on Political Attitudes, Separate Models with Census Tract

Note: Tile color indicates coefficient direction (blue = positive; green with black border = negative); tile size represents the absolute value of the coefficient; and tile opacity reflects statistical significance (opaque = significant; some level of translucent = not significant). Tile size is scaled relative to the largest coefficient in any of the models run. Model results and sample sizes can be found in Supplementary Tables A-4–A-9.

Several patterns emerge across both racial context and racial group. First, if we use geographic racial context—the top panel of Figure 4—as our primary measure of racial context, we would conclude that racial context rarely has an impact on political beliefs. All else equal, we find a significant effect in just three models out of 24. When Asian Americans live in a census tract with a greater proportion of Asian residents, they are more likely to agree that the government should increase deportations of undocumented immigrants, while Black Americans who live in a census tract with more Black residents express stronger feelings of external political efficacy. Previous work has found that Asian Americans are much less supportive of a path to citizenship for the undocumented than Black or Latino Americans (Tran and Warikoo Reference Tran and Warikoo2021), which could explain why Asians who live among a higher proportion of other Asians are less supportive of undocumented immigrants. Additionally, white Americans who live in a census tract with a larger population of whites are more likely to agree that police in their community use the correct amount of force. So, while prior scholarship has argued that geographic community is “a site of coidentification and political action for marginalized groups” (Nuamah and Ogorzalek Reference Nuamah and Ogorzalek2021), our findings demonstrate that, in the outcomes we examine, geographic context rarely has direct political effects.

The middle panel of Figure 4 shows that social context has a relatively larger effect on the attitudes we measure, significantly predicting at least one political outcome for all four racial groups. But the overall relationship is moderate and concentrated among Latino Americans and, to a lesser extent, Black Americans. For Latinos, increasing the in-group proportion in one’s close social network is associated with an increased likelihood to identify as a Democrat; greater opposition to increasing deportations of undocumented immigrants; an increase in opposition to the idea that it is better if English is the only language used in public schools; and a rise in support for federal redistribution to the poor. Black Americans with a larger proportion of close Black social ties similarly have a more Democratic orientation and increased support for federal redistribution to the poor, as well as greater disagreement with the notion that police use appropriate force where they live.

Moving to the bottom panel of Figure 4, we find that psychological racial context—the degree to which someone feels “close” to their racial in-group—is extremely influential for Americans’ political beliefs and policy attitudes. In particular, psychological context is statistically significant for minoritized groups, especially Blacks and Latinos, for whom a stronger feeling of closeness to other racial in-group members is positively associated with more progressive stances on most or all of the issues we examine. For Asian Americans, the effect is less widespread, but closer feelings to other Asians are associated with lower feelings of political efficacy, a more Democratic orientation, and a belief that police use appropriate force. For white Americans, those who feel more psychologically tied to their racial group have more conservative immigration and policing attitudes. Taken together, Figure 4 shows that 15 of the 24 psychological embeddedness coefficients are statistically significant, compared to 10 of the social embeddedness coefficients and 3 of the geographic embeddedness coefficients.

The very different stories told by each panel of Figure 4 highlight why understanding the degree of nestedness between geographic, social, and psychological racial contexts is essential to understanding the impact of racial context on political outcomes. In a highly nested world where each type of racial context overlaps considerably, treating these different measures of racial context as interchangeable makes sense. However, in the less nested world in which we live, the effect of a single type of racial context on political outcomes tells us very little about the effect of a different type of racial context. Figure 4 clarifies this complexity and underscores that understanding the political consequences of racial context requires conceptual and empirical specificity about what type of context is being studied and who is being shaped by that context.

Our analyses so far show the unique relationship between each level of context and a set of outcome measures, but the fact that so many Americans live in cross-cutting racial contexts suggests there may also be interactive effects. How does someone who is simultaneously immersed in their in-group at the geographic, social and psychological levels compare to an individual with cross-cutting contexts—for instance, who is surrounded geographically by their in-group but measures low on social or psychological in-group embeddedness?

To explore the intersection of contexts, we again run a series of regressions. We use the high-low cutoffs employed in Figure 3 to code survey respondents into one of eight categories: highly embedded in their in-group in all three contexts; highly embedded in their in-group in just two contexts (three permutations); highly embedded in their in-group in just one context (three permutations); and finally, not highly embedded in any of the three context measures. We measure the relationship between these eight contextual variables, treated as indicator variables, and each dependent variable. We include the same individual and tract-level controls from Figure 4 and run regressions separately for each racial group.

Table 2 shows the results for one outcome variable—attitudes about immigration—while Supplementary Tables A-10–A-15 show the full results for all outcomes. Individuals who are highly embedded in their in-group on all three contextual levels serve as the suppressed category. The table reveals that, indeed, Americans living in cross-cutting racial contexts can have different attitudes toward racialized policies than those fully embedded in their in-group.

Table 2. OLS Regression of Attitudes about Immigration on Racial Context Conflict

Note: Supplementary Table A-10 provides full regression results. Supplementary Tables A-11– A-15 provide regression results for the other dependent variables. † $ p<0.1 $ , * $ p<0.05 $ , ** $ p<0.01 $ , *** $ p<0.001 $ .

Let us take for starters the results for Latinos, the group for whom federal deportations are often seen as substantively most related to in-group interests. The results show that Latinos who are highly embedded in their racial in-group at the geographic, social, and psychological levels are, on average, more supportive of decreasing federal deportations than similarly situated individuals from the other three racial groups (0.938). However, changing one or two contextual levels from high to low embeddedness can depress this support. Latinos who are highly embedded in their in-group at both the social and geographic levels, but not the psychological level, show decreased support ( $ -0.165, $ $ p<0.05 $ ). This support further declines for Latinos who are weakly embedded in their group at both the psychological and social levels ( $ -0.206, $ $ p<0.05 $ ), but being low in both psychological and geographic or both social and geographic have no effect.

We can compare these results to white Americans, the nation’s majority racial group. Decreasing whites’ in-group embeddedness in one singular context has no significant effect on support for deportations compared to white Americans who are highly embedded in their group in all three contexts. However, having low embeddedness in two contexts begins to liberalize whites. When white Americans experience more out-group diversity in their social and geographic contexts, they are more opposed to federal deportations (0.147, $ p<0.05 $ ); the same is true for those who have social groups that are predominantly white but are low on psychological and geographic embeddedness (0.107, $ p<0.05 $ ).

The results for Black Americans look more similar to those of Latinos than whites: low psychological in-group embeddedness decreases support for the liberal immigration policy position, as does being low in both psychological and geographic in-group embeddedness. For Asian Americans, shifting from completely nested racial contexts to cross-pressured contexts has no significant relationship to attitudes.

Across other outcome variables, we similarly find that—compared to those who are highly embedded in all three racial contexts—experiencing low embeddedness in one or two contexts generally correlates with more liberal attitudes for whites and more conservative attitudes for Blacks and Latinos. These results further accentuate that the effect of racial context on political outcomes is highly variable depending on the context or group being examined. Embeddedness in different levels of racial context can have interactive effects on political views. It is imperative that scholars account for multiple measures of racial context in their studies before drawing broad conclusions about the effect of “racial context” on a given political outcome.

THE “CORRECT” MEASURE OF CONTEXT IS CONTEXTUAL

Citizens are politically affected by the communities in which they are embedded. Yet, as the United States continues to diversify racially and ethnically, and technological advancements expand the nature of our interactions and attachments, scholars must broaden their approach to assessing the sociopolitical effects of racial context. By measuring individuals’ racial group embeddedness in multiple ways—geographically, socially, and psychologically—our work maps and disentangles the effects of racial context.

Wong et al. (Reference Wong, Bowers, Rubenson, Fredrickson and Rundlett2020, 165) posit that “the relationships between contextual predictors and policy preferences could vary a great deal, depending on whether objective or subjective measures are used.” Here, we provide robust empirical support for this idea. Our findings show when and how different types of racial context relate to the political orientations and beliefs of Asian, Black, Latino, and white Americans. We see in these findings that social and psychological racial contexts have more political influence than geographic context—at least on the outcomes we have explored.

It is possible geographic context matters more for other measures including voter turnout or prejudicial attitudes. We take these findings not as evidence that geographic racial context never matters, but rather that other forms of context often matter even when the racial composition of one’s geographic neighborhood does not. In many surveys, geographic context indicators merged with census data are the only contextual variables available—providing low hanging fruit for scholars to consider contextual effects. But these measures are insufficient for capturing how racial embeddedness in one’s group either at the social or psychological level is in fact related to political attitudes, often to a larger degree than more conventional measures like age, gender, and income. Combined with the fact that many individuals sit in cross-cutting racial contexts, using geography alone to proxy for social or psychological racial embeddedness will miss important ways these alternative conceptualizations—and the interactions between them—are in fact related to political attitudes.

Our work points to avenues for future research. Although our focus here is on racial context, our typology of contextual environments—geographic, social, and psychological—is broadly applicable to other social identity categories, such as gender, religion, partisanship, and social class. Here, too, little is known about cross-contextual nestedness and how environments of various kinds influence political attitudes and behaviors. Do Americans also face cross-cutting class contexts? Are they more segregated at the level of social environment than geographic environment when it comes to party? Referring back to Figure 1, our findings show that racial contexts appear more cross-cutting than nested for the average American. But these other contexts of import may fall more to the nested side of the spectrum. Our measures and tests can be adapted for a fuller range of groups and contexts.

It is also important to acknowledge that our measure of geographic context captures where respondents currently live. Unfortunately, our survey lacks information about where respondents lived at other periods of their life and how long they have been in their present location. It is possible that other areas in which respondents resided were more formative in shaping their politics, given research suggesting that other geographic contexts—such as where people grew up and attended college—can be important for political socialization (e.g., Goldman and Hopkins Reference Goldman and Hopkins2020; Raychaudhuri Reference Raychaudhuri2018). We hope that future work may build off our findings by inquiring about geographic context at multiple time periods; doing so would enable a more definitive assessment of the role (or lack thereof) geographic racial context plays in shaping attitudes and behavior, relative to other types of racial contexts.

Similarly, we encourage future work to consider differences between objective and subjective measures of contexts across all three levels we have outlined. Past works have done so already in the realm of geographic context (Velez and Wong Reference Velez and Wong2017; Wong et al. Reference Wong, Bowers, Williams and Simmons2012). These scholars compare how an objective measure of the geographic dispersion of peoples (e.g., tract-level demographics measured through a census) relates to subjective measures of the geographic dispersion of peoples (e.g., survey respondents’ reports on the racial composition of their neighborhoods). Similar work comparing subjective measures of social or psychological context to objective measures are ripe for exploration.

Like V. O. Key, much of political behavior research to date has focused on how two groups—Black and white Americans—react to changes in their racial context. But as the U.S. population becomes increasingly heterogeneous, scholars must broaden their studies to other ethnoracial groups. Our uniquely diverse sample helps add necessary clarity to a large literature on racial context by showing how the relationship between contexts and political phenomena vary across Asian, Black, Latino, and white Americans. We emphasize that when studies are restricted to a single ethnoracial group, the conclusions drawn are incomplete. As such, we urge political scientists to expand their theories, samples, and analyses to include participants from different ethnoracial backgrounds; doing so enhances the validity and generalizability of findings to the broader U.S. population.

Additionally, we note that in a quickly diversifying America, “in-group” and “out-group” have become more complicated (e.g., Pérez, Deichert, and Engelhardt Reference Pérez, Deichert and Engelhardt2019; Smith Reference Smith2014). Our research has focused on the racial in-group side of this equation, showing that Asian, Black, Latino, and white Americans alike are more likely to face cross-cutting contextual forces than they are to experience contextual nestedness. For white Americans, segregation is strongest at the level of social embeddedness; for non-white Americans, co-racial psychological context predominates. Still to explore is the out-group side of our distributions. Allport (Reference Allport1955) argued that, under certain conditions—including equal status, common goals, intergroup cooperation, and support of authorities, laws, or customs—prejudice is reduced when there is interpersonal contact between groups. For example, recent work by Berinsky et al. (Reference Berinsky, Karpowitz, Peng, Rodden and Wong2023) has shown that white missionaries who were assigned to communities with larger Latino populations became more tolerant toward undocumented immigrants, thus demonstrating that direct exposure to racial out-groups in one’s geographic environment can significantly affect immigration attitudes. However, group size also matters: as the size of an out-group increases, interpersonal contact may also increase prejudice toward that group (Fouka and Tabellini Reference Fouka and Tabellini2022). How homogeneous or diverse is the out-group Americans interact with on each of the three contextual racial dimensions, and what is the effect of this on political behavior?

In American politics today, partisan and racial divisions are stronger than ever (Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2018). Given this highly polarized landscape, understanding the contextual conditions under which intergroup contact can decrease antagonism and promote mutual understanding is necessary for helping bridge racial and political divisions. Our study serves as an important step in this effort, by clarifying the concept of racial context and showing when, how, and for whom contextual measures predict racialized political outcomes.