Where do we find interracial couples in the contemporary U.S.? Research on different-sex couples has shown us that the social and demographic context of communities is related to the prevalence of interracial couples (Blau, 1977; Blau et al., 1982; Cready & Saenz, 1997; Fitzpatrick & Hwang, 1992; Hwang et al., 1994). Demographic and economic opportunities in the local area, levels of integration, patterns of migration, norms, and policies governing relationships are all related to intermarriage rates (Torngren et al., 2016). However, most studies on interracial relationships focus on different-sex married couples, raising the question of how the growing prevalence of same-sex and/or cohabiting couples might alter these relationships (Rosenfeld, 2007). Same-sex pairings are on the rise, and same-sex unmarried partners have the highest rate of interracial pairings of all relationship types (Lofquist et al., 2012; Rosenfeld & Kim, 2005). Cohabiting relationships are also more likely to be interracial than marriages (Joyner & Kao, 2005). Without including same-sex and cohabiting couples in our investigations, we cannot fully understand the spaces where we find disproportionate shares of interracial couples today.

We focus on the city context in which couples live because many of the social and economic transactions that govern daily life take place within one’s town or city, providing a potential source of diversity, commonality, and mutual support (Rosenfeld, 2007). Cities may shape the opportunities that make a new relationship possible, and couples also have the agency to choose to migrate to a more supportive city. The prevalence of non-normative relationships (e.g., same-sex, interracial, cohabitating) in an area might also be shaped by contextual mechanisms; however, there are disputes about how much context matters for non-normative relationship prevalence. For instance, previous research has encouraged researchers to consider the impact of political and religious climate on GLB (gay, lesbian, and bisexual) individuals and group distribution (Oswald et al., 2010). However, research has also indicated that areas once deemed “unwelcoming” now have many GLB residents, possibly reflecting a more tolerant social climate (Black et al., 2000; Rosenfeld & Kim, 2005; Kazyak, 2012). Therefore, testing if and how contextual factors influence the distribution (and possible formation) of non-normative relationships, especially those not well studied in the past, adds to our understanding of how contextual factors (which are often tied to public attitudes) may influence geographic variation.

These findings may also give greater insights into social boundaries. Previous research on interracial relationships suggests that the U.S.’s growing number of interracial relationships indicate increasingly porous intergroup boundaries (Lee & Bean, 2010; Qian & Lichter, 2007) but there is far less discussion of how these boundaries may vary across places. The growing numbers of interracial relationships have broad implications for the U.S. regarding social, economic, and political factors (e.g., “majority-minority” demographic shift; see Craig et al., 2018; Alba, 2020). Similarly, one could postulate that the growth of other non-normative relationships (e.g., cohabitation, same-sex relationships) could also have similar implications for boundary rigidity.

In this paper, we use 2012–2014 American Community Survey (ACS) data matched to contextual data to test how the demographic and religious context of cities relate to the odds of observing heterogamous couples, including all couple types in our analysis (different-sex and same-sex couples, and married and unmarried couples). We find that contextual factors in a city predict different odds of interracial partnerships among same-sex and different-sex couples for White, African American, and Latinx respondents. To our knowledge, no study has looked at the presence of religious institutions as a structural predictor of the prevalence of heterogamous relationships in the United States, although past studies have found that religious activities and beliefs relate to attitudes about interracial relationships (e.g., Herman & Campbell, 2012; Perry, 2013a). As we will discuss further in the discussion section, this analysis of interracial same-sex cohabitations and marriages before the Obergefell v. Hodges decision in 2015 (that made same-sex marriages legal in all states) allows us to examine same-sex and different-sex relationships at a pivotal and dynamic time in U.S. relationship history.

Background

Why Focus on Interracial Relationships? Race-Making and the Law

Relationships, especially marriages, have been central to how the concept of race has been encoded in the law in the United States. To make certain the idea of racial purity became normalized, Lee & Edmonston (2005) note that the first anti-miscegenation laws were passed in Maryland in the 1660 s banning Blacks and Whites’ union with other states following suit after that. Pascoe (1991) asserts Maryland’s 1664 law banned “free English women” and Negro slaves” which denote both race and sex specific parameters. Pascoe (2009) notes how miscegenation law is functioned as the “legal factory for the defining, producing, and reproducing of the racial categories of the state” (Pascoe, 2009, 9). This emphasis shows the importance of childbearing for these statutes; miscegenation statutes were often explicitly focused on controlling White women’s sexuality and white supremacist ideas about “protecting” the white family. White men, in contrast, were allowed a great deal of informal sexual access to Black women (Pascoe, 1991) in ways that were perceived as less threatening to White supremacy because they were not perceived as a threat to the institution of the White family (Novkov, 2008). Lee & Edmonston (2005) note the function of anti-miscegenation laws centered around reinforcing ideas of racial purity, difference, and separation in order for whites to maintain power and privilege. These miscegenation laws highlight not only how the state was invested in race-making through its anti-miscegenation laws but also how this race-making in the political, economic, and physical realms reinforced the idea of Black inferiority and White superiority.

During Reconstruction, many of the legal methods used to deny African Americans full citizenship and equality under the law were predicated on the idea that race was an undeniable and inherited characteristic, and allowing intermarriage would threaten that conception of race (Novkov, 2008). The last of these “anti-miscegenation” laws were not rendered unconstitutional until 1967, when Loving v. Virginia was decided by the U.S. Supreme Court. As Novkov (2008) notes, however, 1967 did not mark the end of this issue in the U.S. In 2000, Alabama voters removed the text from the Alabama constitution that made intermarriage illegal, but even though this ban had been legally unenforceable since 1967, 40% of Alabama voters nonetheless voted to keep the text banning interracial relationships in their state constitution. Novkov (2008) explains that this constitutional change was so much more controversial than many expected because of the long, complex history of tying white supremacy specifically to the institution of the “White family” and tying conceptions of what “race” is to legally sanction childbearing relationships. Examining patterns of interracial family ties, then, give us key insights into how these racial boundaries, so rigidly and violently maintained in U.S. history, are changing today.

Couple types and interraciality

In contrast with different-sex marriages, both cohabitations and same-sex relationships are more likely to be interracial today (Rosenfeld, 2007; Rosenfeld & Kim, 2005; Schwartz & Graf, 2009). These relationships have very different histories than different-sex marriages; since both are seen as non-normative relationships, both have experienced their own significant sanctions, but both have also been less regulated by the state regarding interracial sex. These differences make it increasingly important to include same-sex and cohabiting couples in the landscape of interracial pairings today.

Same-sex couples are more likely to transgress a range of social norms than different-sex couples, such as homogamy in race, age (Schwartz & Graf, 2009; Verbakel & Kalmijn, 2014), and education (Jepsen & Jepsen, 2002). Interracial same-sex couples have increased in the past twenty years, from 15% in 1990 to 21% in 2010 (Gates, 2012; Rosenfeld, 2007). The population of gay and bisexual men and women is much smaller than heterosexual men and women, which decreases their dating pool (Verbakel & Kalmijn, 2014). Still, a lack of available partners does not explain the greater rates of interracial coupling. Greater availability of gay men and/or lesbians within a metropolitan area are associated with higher rates of interracial unions for both gay men and lesbians (Schwartz & Graf, 2010). As cohabitation becomes more common (Bumpass & Lu, 2000; Bumpass et al., 1991; Lichter et al., 2010), it is also important to test these patterns of heterogamy across relationship types. Based on this literature, we develop our first hypothesis:

H1

Compared to different-sex couples (married and unmarried), the odds of being an interracial couple will be higher for same-sex couples (married and unmarried).

As we noted above, the city context could be important for these relationships, so it would be logical to also test a hypothesis focused on the relationship between same-sex couple type and city context. Unfortunately, because of the small number of same-sex couples and interracial couples, we only have the statistical power to test this when comparing couples that include at least one White partner.

Demographic Characteristics of Cities

How do structural opportunities affect the likelihood of couples crossing racial boundaries? About 90% of U.S. residents live in metropolitan and micropolitan areas (Core-Based Statistical Areas, or CBSAs; OMB, 2000), but same-sex couples on average live in more diverse, urbanized, and amenity-rich cities than different-sex couples (Baumle et al., 2009; Black et al., 2000; Gates & Ost, 2004). Living in more diverse spaces provides an opportunity for interracial contact, so this might explain greater rates of interraciality.

The association between racial and ethnic heterogeneity in a city and interracial relationships is mixed. Heterogeneous spaces are generally associated with racial/ethnic intermarriage (Blau et al., 1982; Lee & Bean, 2010), but some studies report a non-significant or negative association between city diversity and racial/ethnic intermarriage (Cready & Saenz, 1997; Hwang et al., 1997). Of course, some heterogeneous spaces are very segregated, and residential segregation is negatively associated with intermarriage (Anderson & Saenz, 1994; Peach, 1980). Living in segregated spaces likely limits opportunities for contact with others of a different racial/ethnic group (Anderson & Saenz, 1994; Peach, 1980).

Unlike heterogeneity, the relationship between minority group size and intermarriage is consistent; more racial/ethnic minorities in an area decreases the likelihood of interracial marriages. Larger minority group size likely decreases intergroup interaction, either because of the availability of more intragroup contact for racial minorities or increased perceptions of competition and group threat (Anderson & Saenz, 1994; Blau et al., 1982; Fitzpatrick & Hwang, 1992).

Heterogeneity and group size structure are opportunities for forming interracial relationships (structural factors) and relate to the likelihood that the city will be a supportive environment for interracial couples (normative factors; Torngren et al., 2016). Therefore, we include measures of minority group size and other measures of city diversity such as the proportion of the city that is foreign-born in our models predicting interracial partnerships; that as the size of the Black or Latinx population increases, the odds of interracial partnerships will be lower for Black or Latinx adults, but higher for White adults, as past research has found. We also add to this literature by testing whether racial segregation also creates an environment less conducive to interracial relationships:

H2

Higher levels of segregation in a city will be associated with lower odds of interracial relationships (of all types) for all racial groups.

Evangelical Protestantism, Interraciality, and the Contextual Marketplace

Religiosity is negatively related to interracial relationships (Herman & Campbell, 2012) as well as same-sex relationships (Olson et al., 2006), especially among conservative Christians (Lichterman et al., 2009). This paper focuses on Evangelical Protestants because of findings from previous research regarding religiosity, same-sex relationships, and race. A smaller percentage of Evangelical Protestants than Mainline Protestants and Catholics support same-sex relationships (Perry, 2013b), and Evangelical Protestants are, on average, less open to engaging in an interracial relationship than individuals that are religiously unaffiliated (Perry, 2014). Evangelical Protestants are especially likely to rely on individualistic and cultural (deficiency) explanations of racial inequality rather than structural explanations like discrimination (Emerson & Smith, 2000; Emerson et al., 1999), relying more heavily on colorblindness across multiple contexts. Previous research has also suggested that a long history of racism within American Protestantism (Lichterman et al., 2009) and theologically closed networks (i.e., a lack of connection to the broader community that may impact race relations; see Blanchard, 2007 and Putnam, 2000) as potential factors that shape racial attitudes among White conservative Christians. Moore and Ovadia (2006) also note that local areas with higher concentrations of Evangelical Protestants have lower levels of tolerance (measured as a willingness to extend civil liberties to everyone). Therefore, a higher concentration of Evangelical Protestant adherents (those who identify with, but of course may not regularly attend, an Evangelical Protestant church) in a city may be negatively associated with interraciality (among same-sex and different-sex couples). Religious communities are also negatively associated with county-level trends in cohabitation (Gault-Sherman & Draper, 2012). Considering all of this evidence, we hypothesize:

H3

Higher prevalence of Evangelical Protestant adherence in an area will be associated with lower odds of interracial relationships of all types.

data

Data are from the 2012–2014 IPUMS-ACS (Ruggles et al., 2015), 2010 Decennial Census, and the U.S. Religious Census: 2010 Religious Congregations and Membership Study Metropolitan file (hereafter, RCMS) (Grammich, 2012). We use these years of ACS data because they were measured close in time to the 2010 RCMS, and they are the earliest ACS files to directly measure married same-sex couples, which was not possible in previous waves. We match couple-level data from the ACS to aggregated data from the 2010 Census to examine the demographics of Core-Based Statistical Areas (micropolitan and metropolitan areas greater than 10,000 individuals). We then match this dataset to aggregate RCMS data to measure religious adherence rates in each area.

Identifying Same-Sex and Different-Sex Couples

To identify all unmarried partnerships and married couples, we selected all householders who had a spouse or unmarried partner and matched them with their spouse or partner. In 2013 and 2014, the process is identical for same-sex and different-sex couples. For same-sex marriages, the process is different in 2012 because the Census Bureau recoded married same-sex spouses as “unmarried partners.” We identify same-sex marriages that year using the allocation flag “QRELATE.” These procedures resulted in an initial sample of N = 2,050,737 couples.

As others have noted, however, because their numbers are small, even a small number of errors in the classification of relationship type or the gender of the relationship partners can have huge consequences for the estimation of values for the population of same-sex couples (Lofquist & Lewis, 2015). For that reason, we followed the procedure outlined in Gates (2015), dropping both same-sex and different-sex couples if: (1) the sex of either partner was allocated; (2) the marital status of either partner was allocated; or (3) married partners do not report the same year of marriage. In addition, for same-sex couples only, we drop those who (4) report being married before 2004 (when Massachusetts became the first state to issue same-sex marriage licenses). This undoubtedly results in the exclusion of some couples who are married to a same-sex partner (for example, couples who simply misreport their date of marriage, or who married outside the United States before 2004, etc.). Still, it ensures that our estimates exclude the cases that are most likely to be errors. This procedure means we drop 133,508 married different-sex couples and 14,396 same-sex couples (unmarried and married) from 2013 to 2014 (65,781 married different-sex, 2230 married same-sex and 4829 cohabitating same-sex couples in 2013; 67,727 married different-sex, 2241 married same-sex, and 5096 cohabitating same-sex couples in 2014). Given the recoding of same-sex married couples in 2012, we do not drop same-sex married couples in the 2012 data whose marital status was allocated or who reported different years of marriage because both of these variables were affected by the Bureau’s reclassification. Dropping different-sex couples with allocated marital status or different years of marriage and dropping all couples with allocated sex or same-sex couples who report marriage before 2004 resulted in the loss of 51,216 different-sex couples (48,168 different-sex married couples) and the loss of 895 same-sex couples in 2012. After dropping all of the excluded cases (about 10% of the cases), we have a sample of N = 1,850,722 couples (see Fig. 1). Because 2012 and 2013–2014 data were not coded the same way, we did sensitivity tests using only data from 2013 to 2014 and did not find meaningful differences in the results.

Fig. 1
figure 1

Sample exlusions and listwise deletion

Full size image

Other Sample Restrictions

We excluded couples with individuals younger than age 25 (who may be still completing their schooling) (N = 35,619). We keep individuals up to age 50, retaining both young and middle-aged couples. Excluding those over 50 (N = 619,836) provides a large enough sample size to include same-sex couples while still allowing us to focus on just a few cohorts. Individuals older than 50 are also more likely to be in remarriages, which are typically less racially and educationally endogamous than first marriages. As in many studies of intermarriage, we excluded foreign-born individuals who immigrated after age 13 (N = 233,119) because these couples are more likely to have married abroad or, because they attended a significant amount of schooling outside the U.S., to have been heavily influenced by union formation norms in their country of origin, which is distinct from our research questions here (Qian & Lichter, 2007, 2011). People who identified as “other race” were excluded because of the small sample size. Those who identified as multiracial were also excluded (because of small sample size and the difficulty in establishing whether or not a partnership is seen by the couple as interracial if a partner shares one but not all of their racial identifications; N = 13,692). We excluded couples with any missing data on a partner’s race/ethnicity and residential location. This leaves us with a final sample of 354,346 marriages and 59,345 cohabitations. We handled missing data through listwise deletion.

Race and Ethnicity

We classify each partner into mutually exclusive racial and ethnic groups. Those who identify as Latinx are placed in a Latinx category (N = 27,241), without regard to what they selected on the second question. The non-Latinx racial categories are White (N = 307,698), Black (N = 24,125), American Indian/Alaska Native (N = 711), and Asian/Pacific Islander (N = 6,280). We then code same-sex and different-sex couples as interracial (e.g., White/Asian) or same-race (e.g., White/White) dyads. Most same-sex couple types have small sample sizes; thus, we consider only the largest groups: same-race couples that are White, Black, Asian, or Latinx, and interracial couples that are White/Black, White/Asian, and White/Latinx.

Couple-Level Control Variables

As noted above, we distinguish same-sex relationships from different-sex relationships and marriages from cohabitations. We include couples’ average age and the absolute difference in the partners’ educational attainment. We collapse each partner’s years of completed schooling into six categories: (0) no schooling, (1) no high school diploma, (2) diploma/GED, (3) some college, (4) bachelor’s degree, and (5) master’s degree and higher. Then, we compute the difference between the partners, testing for educational heterogamy. We also tested a control for whether the respondents had moved within the last year; about 2% of the sample had moved from outside the state, and about 8% had moved within the state. Including this variable decreased our sample size considerably but did not change the relationships shown here, so we have excluded it. Our tests showed that couples who had moved in the last year were more likely to be in interracial relationships compared to White or Latinx homogamous relationships.

Contextual (CBSA-Level) Variables

We focus on the city context in which couples live because many of the social and economic transactions that govern daily life take place within one’s city (Rosenfeld, 2007). Cities may shape the opportunities that make a new relationship possible, and couples also have the agency to choose to migrate to a more supportive city. When interpreting these relationships, it is important to remember that where one lives may not be independent of in-group preferences formed earlier in life. For instance, an individual may move to San Francisco because they prefer to live in a diverse city, and that same preference may also shape their dating patterns.

We use contextual-level measures of the CBSA context as predictors of intergroup contact: racial group size, immigrant group size,Footnote 1 racial/ethnic residential segregation, rates of Evangelical Protestant adherence, city size, proportion highly educated, and region. We also consider interactions between these contextual variables and whether or not the couple is same-sex to test for differences in the importance of these structural factors across couple types, as we know little about how these mechanisms might operate differently for same-sex couples.

Group Size and Segregation

We calculate the proportion of the CBSA that is Black, Latinx, and other minority groups (i.e., Asian, American Indian/Alaska Native, Native Hawaiian/Other Pacific Islander, some other race, and two or more races combined). We take the natural log of the last two measures because of the skewed distribution of the variable. We also control for the proportion foreign-born in the CBSA to test whether a large immigrant community within a CBSA may heighten group threat, thus increasing social distance (Lee & Bean, 2010), or conversely, encourage more exogamy among some native-born groups (Campbell & Martin, 2015). We use the index of dissimilarity to measure White/Black and White/Latinx segregation. This familiar index measures the minimum proportion of a group that would need to change neighborhoods (census tracts) to establish an even distribution of both groups across all tracts within a CBSA (Massey & Denton, 1988).

Rates of Evangelical Adherence

We use the natural log of Evangelical Protestant adherence per 1000 persons from the 2010 RCMS, which contains religious adherent data at the metropolitan level (Grammich, 2012). An “adherent” is anyone who self-identifies with an Evangelical denomination (who may or may not belong to a church). We do not include Mainline Protestants and other denominations because Evangelical Protestants are more likely to oppose racial intermarriage and same-sex marriage (Olson et al., 2006; Perry, 2013a, 2013b).

Proportion Highly Educated

We measure the proportion of adults aged 25 and over in the CBSA who have completed a bachelor’s degree or more. In the aggregate, we anticipate that the more educated an area, the greater the tolerance for non-normative relationships. Interracial relationships are significantly more common among the college-educated (Wright et al., 2003) and the relationship between higher education and lower average levels of racial prejudice has been increasing over cohorts (Quillian, 1996). Local areas with higher average levels of education also express greater social tolerance generally (Moore & Ovadia, 2006).

Total Population and Region

We control for the natural log of city population size because interracial couples are disproportionately in large cities (Baumle et al., 2009). We also control for living in the West, with its higher rates of intermarriage (Choi & Tienda, 2016).

Method

We treat the racial makeup of the couple as the dependent variable in a multinomial logit framework and test how couple-level and contextual-level characteristics predict the likelihood of different couple types (e.g., White/White versus White/Black couples). Our first model estimates the relationship between the city characteristics and whether couples are homogamous White couples (the reference group), White/Black, White/Latinx, or White/Asian couples. We conduct a similar comparison of homogamous Black to Black/White and Black/Latinx couples, and finally, we compare homogamous Latinx to Latinx/White and Latinx/Black couples. Because the data are nested (couples within CBSAs), we estimate multinomial logit models with adjustments for clustering to obtain estimates and standard errors that are not biased by the non-independence of cases within the N = 256 CBSAs. In the table with the largest sample size (comparing couples with at least one White partner), we also include a cross-level interaction between the same-sex status of the couple and the level-2 (CBSA) variables, allowing us to test whether or not the relationship between contextual characteristics of the area and interracial partnerships varies across different-sex and same-sex couples. Using “svy” command in Stata, we weighted the results with an ACS level-1 household weight divided by three (because we are aggregating three years of data) and incorporated the replicate weights (Ruggles et al., 2015).

Results

Table 1 provides summary statistics for the couple-level and CBSA-level characteristics of married and cohabiting different-sex and same-sex couples. Cohabiting couples are more likely to have partners of different races than married couples (Jepsen & Jepsen, 2002; Rosenfeld, 2007). Among the various interracial couple combinations, White/Latinx couples are the largest, followed by White/Asian couples and White/Black couples. On average, same-sex couples live in more highly educated areas than different-sex couples (Rosenfeld, 2007). Additionally, same-sex married couples are more likely to live in CBSAs with lower rates of Evangelical Protestant adherence and higher proportions of foreign-born residents compared with different-sex cohabitating couples.

Table 1 Sample Descriptives, by Couple Type
Full size table

Multinomial Logit Models: Comparisons with Homogamous White couples

For Table 2, we begin with couples with at least one White partner (with White/White couples as the reference), giving us the largest sample and the greatest statistical power. We tested cross-level interactions between each of the level-2 variables and the level-1 indicator of same-sex couples to test whether the relationship between context and exogamy varied by couple type. Table 2 only includes the interactions that were robust across different specifications: the rate of Evangelical Protestant adherence, racial composition, segregation, and region.

Table 2 Multinomial logit models (reference = White/White couples)
Full size table

Couples who are married, older, or have smaller educational disparities are less likely to be in an interracial relationship. Contextually, a city with a higher proportion of immigrants is positively associated with being in a White/Latinx and a White/Asian relationship. In addition, cities with a greater proportion of individuals with a Bachelor’s degree are associated with White/Asian coupling.

Focusing on the interactions, proportion Latinx and same-sex couples are positively associated with White/Black relationships. A higher proportion of other minorities and same-sex couples are positively associated with White/Asian relationships. Same-sex White/Black coupling also occurs at lower rates in the West. We also expected that higher levels of segregation in a city would be associated with lower odds of an interracial relationship—we do not find support for this in our three models. The test of our final hypothesis showed that Evangelical adherence is significantly associated with lower White/Black same-sex relationships. Figure 2 illustrates these relationships. Greater Evangelical adherence decreases the log odds of all same-sex interracial couple types, most dramatically among Black-White relationships, while it increases the log odds of White/Black and White/Asian different-sex couples. This suggests a varying degree of tolerance for nontraditional unions in cities with many Evangelicals: more tolerance toward interraciality compared to same-sex relationships, as we might expect.

Fig. 2
figure 2

Predicted probability of couple types, by Evangelical Protestant adherence in the CBSA (from Table 2)

Full size image

Multinomial Logit Models: Comparisons with Homogamous Black Couples

In Table 3, we examine couples with at least one Black partner, where Black/Black couples are the reference. The same-sex indicator cannot be included as a control because of low statistical power, but the model still includes both different-sex and same-sex couples. We again observe that being married and older decreases the log odds of being in an interracial relationship.

Table 3 Multinomial logit models (reference = Black-Black couples)
Full size table

At the contextual level, increases in CBSA proportion Black significantly decrease the log odds of observing both Black/Latinx and Black/White couples, and increases in the Latinx population are associated with greater odds of observing Black/Latinx couples. This supports our hypothesis that larger minority groups will have more in-group partnerships, and they are also more likely to partner with other groups if they are larger. There is a significant negative relationship between the share of the CBSA that is foreign-born and couples with a Black partner in a more educated CBSAs have more Black/White relationships, and living in the West has a positive relationship with the log odds of Black/White and Black/Latinx couples.

Turning to segregation, we observe that living in a city with more White-Black segregation has a significant negative relationship with the log odds of interracial unions, which supports our hypothesis, and suggests that segregation is especially important for the color line between African Americans and other groups. Finally, as predicted, when there is a high rate of Evangelical adherence in a CBSA, couples with at least one Black partner are especially likely to be homogamous, which is consistent with the literature suggesting that religiosity is particularly strongly related to negative beliefs about relationships with African Americans (Herman & Campbell, 2012).

Multinomial Logit Models: Comparisons with Homogamous Latinx Couples

In Table 4, homogamous Latinx couples are the reference group. The couple’s average age is positively associated with interraciality, consistent with work that suggests that interracial relationships are declining slightly for recent cohorts of Latinxs as the group size grows (Campbell & Martin, 2015). We again find the inverse relationship between group size and interracial relationships for Latinxs, as well as the significant positive relationship between Black group size and Latinx/Black couples. Areas with more foreign-born residents also have fewer Latinx/White and Latinx/Black relationships, suggesting stronger boundaries in places with more immigrants. White-Latinx segregation has the hypothesized negative relationship with Latinx/White unions (Anderson & Saenz, 1994). Finally, the Evangelical adherence rate has a significant and negative relationship with Latinx/Black relationships, emphasizing the importance of this city characteristic for the interracial relationships among African Americans.

Table 4 Multinomial logit models (reference = Latinx-Latinx couples)
Full size table

Discussion and Conclusions

We test whether demographic and religious contexts correlate with interracial relationships. Each of our hypotheses is at least partially supported for White, Black, and Latinx partners in different-sex and same-sex couples. With respect to our first hypothesis, we find that same-sex couples are more likely to be in White/Black interracial pairings specifically. That is, after controls for demographics and context, we observe no difference for the weaker boundaries between White, Asian, and Latinx partners, but the stronger boundary between White and Black partners is crossed more often by same-sex couples. Our research shows that crossing racial lines with someone who identifies as “Black” is still less likely than other racial groups, which echoes the long history of white supremacy and race-making in the United States. The social problem created by white supremacy, an ethos of anti-Blackness, solidified through law and culture, defined flexible racial boundaries that allow some to cross.

Our second hypothesis has more mixed results: greater segregation usually, but not always, is associated with fewer interracial relationships. As expected, we also found that larger populations of color increase the rates of interracial relationships for White partners, but larger Black and Latinx populations also increased rates of same-race partnering for these groups. This is to some extent demographically anticipated; same-race relationships are easier to form when the group is large, because of the greater availability of potential partners, but also perhaps because of the greater ability to follow group solidarity preferences. Larger groups of color also increased the likelihood of Black-Latinx partnerships. These findings are important to keep in mind when we contextualize the history of anti-miscegenation laws in the United States, which lasted the longest and were the strictest in the region where the Black population was the largest (in the South) (Novkov, 2008). Today, racial prejudice remains greatest in areas with large populations of color (Quillian, 1996). Still, the countervailing influence of greater contact between racial groups might, over time, change the meaning of some of these racial boundaries. As Novkov (2008) noted, it was the threat of interracial families and childrearing that most directly motivated many of the laws underpinning White supremacy.

Our final hypothesis concerns the relationship between the rate of Evangelical Protestant adherence in a city and interracial unions, and we find a clear relationship with reduced rates of White/Black same-sex couples. We suspect fewer couples live in these areas because relationships with African Americans and same-sex relationships are often stigmatized by conservative Christians (Lichterman et al., 2009), and areas with large Evangelical Protestant populations are less tolerant (Moore & Ovadia, 2006). It is apparent religious adherence has a significant role in the social fabric of a city, especially for non-normative relationships.

The limitations of our analysis are, in part, driven by the size of the sample. Only our analysis of unions with White partners included a cross-level interaction between local context and same-sex couples, as this group had a sufficient sample size to provide reliable estimates. Our sample was also constrained because we limited the couples included in order to have more confidence that each same-sex couple was, in fact, same-sex (because small coding errors can have a large impact on the estimates of relationships for smaller samples). This limitation also required us to limit our analysis to unions that include White, Black, and Latinx partners, as these couples had a large enough sample to provide reliable estimates. Lastly, not all couple types are equally likely to live in CBSAs, so we miss some couple types more than others by excluding non-urban and rural areas. Even with these limitations, our findings advance our understanding of the more diverse world of couples.

It is important to keep the historical context in mind. These data were collected just before Obergefell v. Hodges (in 2015) made same-sex marriages legal in all states. Hence, while same-sex marriages were available in parts of the country and to couples willing to travel to marry, they were not recognized in all places. This makes this particular historical moment quite interesting. It is likely that a significant portion of the same-sex cohabitations in our sample would have been same-sex marriages if marriage was equally available to all. It is also likely that the change in the legal status of same-sex marriage did change the attitudes of some individuals and some Evangelical congregations, so we encourage future research to test the persistence of these patterns across contexts. While much press coverage has focused on the truly exceptional increase in overall approval of same-sex marriage in recent decades, it is important to note that local variation in this support persists, and many continue to support discrimination against LGBT individuals (Kaufman & Compton, 2021). Same-sex marriage availability is of more personal and structural importance to some GLB individuals than others. For example, Lee (2018) found that Black GLB respondents were more likely to think of same-sex marriage as an important policy if they were parents.

Our research shows the importance of context for the heterogamy patterns of different-sex and same-sex couples. Relative to different-sex couples, some same-sex couples are more likely to be interracial, and our findings show that Evangelical Protestant adherence and segregation are important contextual variables that have a different relationship with unions for same-sex White/Black couples than for different-sex couples. Future research should delve further into the existence of a Black/non-Black binary among interracial unions as well, given our findings suggesting that certain contexts may strengthen the relationship boundaries around Black partners. Lastly, future research should continue to focus on the impact of the religious context on different unions. Our findings suggest that it is important for different-sex couples, same-sex couples, and couples with a Black partner.