Intimate Partnerships and Racial Ascription: A Multidimensional Approach

The formation of intimate unions is a primary site where race is socially constructed (Burton et al., Reference Burton, Bonilla-Silva, Ray, Buckelew and Freeman2010; Joyner and Kao, Reference Joyner and Kao2005; Kalmijn Reference Kalmijn1998). The racial composition of couples highlights the ways race persistently represents a key criterion for mate selection (Kalmijn Reference Kalmijn1998; Robnett and Feliciano, Reference Robnett and Feliciano2011; Yancey Reference Yancey2009). Despite increasing interracial union formation over time (Fu Reference Fu2010; Qian and Lichter, Reference Qian and Lichter2011), racial endogamy remains a cultural and behavioral norm (Rosenfeld Reference Rosenfeld2008; Steinbugler Reference Steinbugler2012), signaling the durability of boundaries between racial groups (Qian and Lichter, Reference Qian and Lichter2007; Vasquez Reference Vasquez2015). Although attitudes stigmatizing intermarriage have declined significantly (Herman and Campbell, Reference Herman and Campbell2012; Perry Reference Perry2012), marital partners and, to a lesser extent, cohabiting couples are still more likely to match on racial background than education or religion (Rosenfeld Reference Rosenfeld2008; Sassler Reference Sassler2010). In fact, studies of online dating reveal that race is frequently used to narrow the field of potential matches, revealing clear exclusions that reinforce racialized hierarchies through excluding African American and other non-White partners (Lundquist and Lin, Reference Lundquist and Lin2015; Robnett and Feliciano, Reference Robnett and Feliciano2011; Yancey Reference Yancey2009). Patterns of assortative mating and preference underline how seeing and understanding race is central to union formation.

Beyond patterning who we select, norms around racial sameness may shape how we perceive and interpret the race of partners. Prior work reveals that race can be interpreted as either aligning with or diverging from the identity of the perceived (Feliciano and Robnett, Reference Feliciano and Robnett2014). According to the multiple dimensions framework, race is experienced as a series of interconnected dimensions (López et al., Reference López, Vargas, Juarez, Cacari-Stone and Bettez2017; Roth Reference Roth2016; Saperstein and Penner, Reference Saperstein and Penner2012) which segment race as self-identification from race as observed by outsiders (Roth Reference Roth2010; Saperstein Reference Saperstein2006). This framework draws analytical insight from incidences of racial mismatch, or instances in which race as identified and race as observed differ (Campbell and Troyer, Reference Campbell and Troyer2007; Roth Reference Roth2010; Song and Aspinall, Reference Song and Aspinall2012). The possibility of mismatch varies clearly by race/ethnic group. The large majority of White and Black adults are observed as a race that matches their self-report (Penner and Saperstein, Reference Penner and Saperstein2013; Saperstein and Penner, Reference Saperstein and Penner2012), while 10% to as high as 50% of self-reported Latinx individuals, Asians, Middle Easterners, or American Indians are racially identified by a third party (e.g. an interviewer) in ways that do not line up with their reported identities (Boda Reference Boda2018; Song and Aspinall, Reference Song and Aspinall2012; Vargas and Stainback, Reference Vargas and Stainback2016). Patterns of racial mismatch are variable but demonstrate that observed race operates and is read independently from identity.

Given the possibity of mismatches and a context in which romantic racial sameness is a norm, might romantic partners “see” relationships as racially endogamous even if their partner’s racial identification differs? Exploring this requires more nuance in data gathering on racial perception. Indeed, an unspecified level of racial mismatch can occur in day-to-day interactions, but most of what we know leverages survey or administrative data that taps the racial perceptions of a “generalized other”, or anonymous persons we interact with briefly. These patterns are revealed through creative comparisons of racial self-reports with race reports provided by an interviewer (Campbell and Troyer, Reference Campbell and Troyer2007; Laster Pirtle and Brown, Reference Laster, Whitney and Brown2016; Saperstein Reference Saperstein2006), a proxy respondent (Porter et al., Reference Porter, Liebler and Noon2016), health officials (Frost et al., Reference Frost, Taylor and Fries1992; Kressin et al., Reference Kressin, Chang, Hendricks and Kazis2003; West et al., Reference West, Geiger, Greene, Harris, Liu, Barton, Elmore, Rolnick, Nekhlyudov, Altschuler, Herrinton, Fletcher and Emmons2005), or how a respondent believes they are seen by “other Americans” (Vargas Reference Vargas2015). Additionally, racial ascription is explored through studies gauging the interpretation of racially ambiguous phenotypes in experimental settings (Franco and Franco, Reference Franco and Franco2016; Khanna Reference Khanna2010; Rockquemore and Brunsma, Reference Rockquemore and Brunsma2002; Roth Reference Roth2010). Studying perceptions of unknown persons can expose the role of racial stereotypes that are cued in these moments of appraisal and which guide racial ascriptions (Feliciano Reference Feliciano2016; Freeman et al., Reference Freeman, Penner, Saperstein, Scheutz and Ambady2011; Saperstein and Penner, Reference Saperstein and Penner2010, Reference Saperstein and Penner2012). While these assessments have shown that such information can shape how race is perceived, they explore a narrow range of interactions that individuals encounter.

Encounters where individuals are known to each other, by contrast, are under-explored as sites where processes of racial ascription are operating. Little (if any) systematic information has examined the prevalence of racial mismatch between individuals in established relationships. This is surprising, as qualitative studies of racial and ethnic identity formation commonly reference the varied ways friends or even family members perceive race. Racial identity is often conceptualized as inherently interactional, resulting from negotiations between individuals and those who perceive them (Khanna Reference Khanna2010; see also Rockquemore and Brunsma, Reference Rockquemore and Brunsma2002). Language, culture, and history inform the interpretation of racialized phenotypes (Boda Reference Boda2018; Khanna Reference Khanna2010; Lee and Bean, Reference Lee and Bean2010; López et al., Reference López, Vargas, Juarez, Cacari-Stone and Bettez2017; Ocampo Reference Ocampo2016). Outsiders are often invested in viewing race in a particular way, specifically if it confirms stable notions of how race/ethnic categories are composed. Similarly, those who are perceived respond to or even drive these impressions by adopting performative strategies (e.g., “acting Black”, “acting White”) to guide how they desire to be understood (Khanna Reference Khanna2010; see also Rockquemore and Brunsma, Reference Rockquemore and Brunsma2002). Taken together, these works reveal that race continues to be a dynamic concept among people who know each other well.

Constructing Race “Intimately”: Dynamics of Mismatch when Race is Reported with Public Data

The current work explores two different facets of how intimate partnerships may operate as a potential landscape where racial mismatch can occur. First, a partner’s racial appraisal may generate a racial mismatch if they view the race of their partner differently from how that partner identifies (e.g. a person understandinging their partner as Latinx when the partner identifies as White). The social dynamics that generate mismatches may extend to these established relationships. Exploring whether mismatches occur can reveal the ways that social and situational contexts have bearing on racial appraisals, even among individuals that know each other intimately.

Second, these mismatches can be parsed further to signal whether individauls view their partners as racially/ethnically similar to or different from themselves. Prevailing norms of endogamy and racial sameness may shape how these mismatches intersect with the race of the perceiver, specifically towards certain types of racial consistency or inconsistency. An adult may see their partner as Latinx, even as that person identifies as White, if they themselves are Latinx. Appraising this type of consistency points to circumstances where a partner “sees sameness” between themselves and their partner, even when that partner self-identifies as a different race/ethnicity. This is distinct from a racial appraisal that generates inconsistent readings between one’s self and their partner, or “seeing difference”. Social tendency towards coupling within racial groups suggests that should mismatches happen at all, individuals may interpret their partner’s race as consistent with their own racial identity, rather than perceive difference.

Drawing out these dynamics of racial perception for intimate partners can inform how we conceptualize and measure intra- and interracial coupling. The prevalence of interracial romantic unions is considered a “litmus test” of race relations. Conceptually, they signal circumstances in which individuals are willing to engage mutually understood racial differences (Kalmijn Reference Kalmijn1998; Qian and Lichter Reference Qian and Lichter2007; Vasquez-Tokos Reference Vasquez-Tokos2017). Interracial romantic unions, marital or cohabiting, have persistently been highest between Asians and Whites and Hispanics and Whites, and least common between Blacks and Whites, reflecting group-level social distances occurring in a range of domains (Qian and Lichter Reference Qian and Lichter2007, Reference Qian and Lichter2011). These distances become less clear as racial mismatch informs whether a couple would be defined as an interracial couple and how individuals interpret racial differences. For example, a person who understands themselves as “White”, but is seen as “Latinx” by a partner who identifies herself as “Latinx”, would describe their union as interracial, while their partner would not. Conceptually, this means that while one party sees themselves as crossing a racial line, the other party understands that they are adhering to endogamous norms.

Mediating the ways partners report the race of themselves and their partners, however, is the nature of racial and ethnic data collection itself. This collection of information requires individuals to articulate their race within a specific range of categories that are preset, institutionally generated, and monitored (Mora Reference Mora2014; Perez and Hirschman, Reference Perez and Hirschman2009). Such categories or organizations of categories are often in tension with the labels individuals would apply for themselves (Dowling Reference Dowling2014), producing scenarios where individuals often do not “see themselves” on forms. A key example is the querying of “Hispanic origin” as a question that is separate from a question on “race”. While many perceive Hispanic as a racial category, alongside other categories such as “White” and “Black” (Dowling Reference Dowling2014; Frank et al., Reference Frank, Akresh and Lu2010; Golash-Boza and Darity, Reference Golash-Boza and Darity2008; Roth Reference Roth2010, Reference Roth2016), the U.S. Census (and many other social surveys) places reports of Hispanic origin into its own question. These responses intersect with race responses, thus requiring respondents to declare their identification as composed of race and Hispanic-origin (e.g. White–Hispanic, Black–Hispanic, White–Non-Hispanic). Given the possible options in reporting race/ethnicity for Hispanic persons, we anticipate a heightened level of inconsistency between observed and self-reports (i.e. mismatch) when reporting both race and Hispanic origin for one’s partner.

The meaning behind race declarations for Hispanic individuals is debated. While some presume it conveys information on racialized characteristics such as skin color (Frank et al., Reference Frank, Akresh and Lu2010), it may also inform notions of American identity. Notably, many choose to “opt-out” and declare “some other race” on their forms (Dowling Reference Dowling2014; Roth Reference Roth2010). A small, but growing, body of literature attempts to understand the dynamic nature of racial and ethnic categories through the lens of outsiders (Abascal Reference Abascal2020; Roth Reference Roth2018; Wade Reference Wade2005). For example, Maria Abascal (Reference Abascal2020) finds wide variability in whether an individual is classified as “Hispanic” by outsiders and that this perception is related to different contextual characteristics. This work highlights the possibility of divergence between outsiders’ perceptions of “Hispanic” and self-identification as such, while further suggesting that these perceptions are linked to other social characteristics (Abascal Reference Abascal2020; Wade Reference Wade2005). However, these studies rely on the appraisals of strangers, and it is an open question as to whether the same level of ambiguity in identification and perception exists among individuals within an established relationship. Understanding the likelihood that ethnic-racial designations are perceived and validated externally can further unpack how the public understands these designations and the differences between them.

Factors Driving Racial Mismatch within Intimate Unions: Research Hypotheses

We consider various social processes that shape racial appraisals in relationships. We may anticipate at the outset that members of race/ethnic groups who are most likely to be perceived differently than they identify, such as Latinx adults (Vargas and Kingsbury, Reference Vargas and Kingsbury2016; Vargas and Stainback, Reference Vargas and Stainback2016), would be the most likely to experience racial mismatch in intimate unions, while self-identified Whites or Blacks would be least likely to encounter mismatch. This may be particularly common given the broader debates on the way that the measurement of race and ethnicity conflicts with how the public understands these concepts. In the section below, we consider the various factors that may drive or facilitate racial mismatch between intimate partners. We theorize that the stability of their relationship, social status characteristics of the perceiver and perceived, and the investment in perspectives on race-related matters inform whether racial appraisals generate racial mismatch or consistency within an intimate union.

Beyond race of the perceiver or the perceived, we anticipate that aspects of relationships that shape how racial information is presented and understood will drive whether an individual’s report of their partner’s race/ethnicity aligns with their partner’s identification, as well as whether they view this race/ethnic designation as consistent with their own stated identification. We begin with the presumption that the amount of information one receives about their partner is directly related to the amount of time individuals spend with each other or the strength of ties between the partners. As prior explorations of racial mismatch draw from superficial or brief encounters, information that provides insight into one’s race is likely improved with more time spent.

Therefore, the stability or duration of a relationship is likely to affect how race is appraised, with longer relationships or stronger ties leading to more information. Longer or more established relationships (i.e. marital/co-residential) may lend themselves to explicit discussions of racial identity that can facilitate sharper understandings of the way partners would classify themselves. This suggests a degree of intimacy that may (or may not) include an appraisal of race that is more likely to be closer to the stated identity of their partner when compared to shorter or less stable relationships. Intact relationships or relationships of considerable length likely provide a range of information that would result in more alignment between identification and observation. Therefore, we expect that established intimate unions or those that are longer lasting, regardless of their current standing, are less likely to produce a mismatch.

In addition, the relative social status standings of couples may also impact how race is perceived and understood, particularly given how it is presented on forms. As prior work has shown that social status attributes can inform the likelihood of understanding a person as a member of a privileged or stigmatized group (Penner and Saperstein, Reference Penner and Saperstein2013; Stepanikova Reference Stepanikova2010), navigating racial classification of self and others may be patterned across socioeconomic lines. Status characteristics such as being unemployed and having been incarcerated have been shown to shape how race is perceived by others (Penner and Saperstein, Reference Penner and Saperstein2013). Further, given the range of public responses to the presentation of race and Hispanic origin questions (Dowling Reference Dowling2014), familiarity with social institutions that request racial data may lower the likelihood of providing mismatched reports. Beyond familiarity with data, highly educated individuals may be more invested in conversations around race that may depress the likelihood of mismatch. We anticipate that individuals exposed to environments such as paid work, or a university setting, will be more familiar with dynamics of race reporting.

Finally, ideologies or beliefs that appear to support endogamous norms may also impact how race is understood. Such beliefs may produce racial consistency with the observer—that is, individuals may be especially inclined to “see” their surroundings or their relationships as racially homogenous. Similar to the ways racially stereotyped beliefs can inform racial ascription (Freeman et al., Reference Freeman, Pauker, Apfelbaum and Ambady2010), certain attitudes may guide how couples interpret each other’s race. Norms of homogamy are a powerful context for relationship formation (Van Zantvliet and Kalmijn, Reference Zantvliet, Pascale and Kalmijn2013) and persist as a cultural norm within families (Kalmijn Reference Kalmijn1998; Steinbugler Reference Steinbugler2012). Investment in racial homogamy may reflect built-in preferences to see certain attributes or identities that may not be present when relying on the partner’s self-identification. Additionally, strongly identifying with a racial/ethnic group may influence how race/ethnicity is expressed within a romantic relationship. Holding and encountering perspectives that convey race as salient may shape the likelihood of mismatch in a direction that confirms ideological positions.

Data and Sample

Investigating racial mismatch and consistency/inconsistency requires unique sets of data that provide information on one person’s race and Hispanic origin from at least two individuals. As we discuss above, while most surveys and demographic databases provide self-reported race and ethnicity, a share of surveys also gather this information on individuals from sources who observe respondents, usually an interviewer or a proxy for the respondent. However, we draw on one of the only (if not the only) datasets that provides both self-reported and observer reported race and ethnicity information where the observer has an established relational tie to the person they are perceiving. The Fragile Families and Child Wellbeing Study (FFCWS) is a longitudinal dataset of parents of recently born children where both mothers and fathers are surveyed. These data were gathered between 1998 and 2000 within a random selection of hospitals within twenty U.S. cities with populations of 200,000 or more (Reichman et al., Reference Reichman, Teitler, Garfinkel and McLanahan2001). This survey aims to tap experiences of family disadvantage and poverty and, therefore, an oversample of unmarried parents is included, comprising roughly three-quarters of nearly 5000 births (Reichman et al., Reference Reichman, Teitler, Garfinkel and McLanahan2001). In addition, the geographic scope of the sampling frame is unique as sixteen of these cities were selected based on a random sample of U.S. cities with populations over 200,000, and four were included in the study because of distinctive policy and economic environments that cover a range of social and public safety nets.

This dataset is ideally suited for exploring how the multidimensionality of race and Hispanic origin unfolds within intimate unions. These data include couples that are both resident and non-resident, providing a range of relationships where race and Hispanic origin are interpreted. Although the data are approximately twenty years old, the FFCWS is one of the only data sources that queries race, Hispanic origin, and other information of both mothers and fathers and asks mothers questions about her child’s birth father. We acknowledge that this time point precedes important shifts in the collection of race data, such as the inclusion of multiple-race reporting; however, it includes the separation of race and Hispanic origin that is employed currently in federal data collections (Office of Management and Budget 1997). Mothers and fathers are also asked about their race and ethnicity at the time of the baseline wave; they are not asked to re-report at later waves. Although fathers are asked their race and ethnicity at a later wave if they are not reachable at the baseline wave, all of the mother’s reports of the father’s race, the mother’s self-reports, and self-reports from the reachable fathers are asked at baseline. We restrict our analyses to data from the baseline survey.

While 4898 mothers were sampled at the time of the baseline survey, fewer fathers (n=3830) were reachable and agreed to participate in the study (Bendheim-Thoman Center 2008). Inclusion in our analytic sample requires valid information on both mother’s and father’s reports of their own race and Hispanic origin and reports by the mother of the father’s race and Hispanic origin. A total of 167 cases were dropped due to missing information on either of these items. An additional 196 cases were dropped due to missing information on independent variables. We employ listwise deletion to handle missing data, resulting in a final sample size of 3467 pairs of parents.

Variables

Classifying Mismatch: Considering Race and Hispanic Origin

To capture mismatch, we begin with cross-classifications of father’s self-reported race/ethnicity and mother’s report of his race/ethnicity. To account for the various ways in which race and Hispanic origin are rendered in surveys (as well as public discourse), we show two ways this mismatch can occur in Table 1 across two panels. First, we show the prevalence of mismatch where Hispanic ethnicity is situated as a “race” alongside other groups; thus, all reports of fathers as Hispanic, regardless of race reports, are aggregated into one category (see Panel A). Therefore, matches and mismatches in Panel A refer to discordance across a conventional set of race/ethnic categories (e.g. non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian, etc.), with a collapsed Hispanic category.

Table 1.

Mother’s Report and Father’s Report of Father’s Race and Hispanic Origin

¹ Percentages reflect the number of fathers by mothers’ report of their race and Hispanic origin (numerator) among all fathers who self-report this race and Hispanic origin category (denominator). Frequencies in parentheses.

² Percentages reflect the number of Hispanic fathers by mothers’ report of their race (numerator) among Hispanic fathers who self-report this race category (denominator). Frequencies in parentheses.

Source: Fragile Families and Child Wellbeing Study, Baseline Wave

In Panel B, we utilize the full range of detail in race and Hispanic origin to identify matches and mismatches in reports of father’s race/ethnicity. This panel captures differences in reports of the father’s race among fathers who were reported as Hispanic by both mothers and fathers. (If one parent reported the father as Hispanic and the other did not, this mismatch would be captured in Panel A.) Experiences of Latinx groups often differ based on reported race, and these distinctions within the Hispanic category are particularly important to capture, as ideas on what constitutes racial difference among this group are in flux (Dowling Reference Dowling2014; Roth Reference Roth2010). We, therefore, account for the potential that how individuals see race is layered in how partners combine their responses to Hispanic origin and race reports. For example, if a mother reports the father’s race and Hispanic origin as “Hispanic White” and he reports himself as “Hispanic Other”, this is classified as a mismatch. A match occurs when the mother describes her newborn’s father with the same combination of race and Hispanic origin categories.

Turning first to Panel A, we find a great degree of concordance between mothers’ reports of fathers and the father’s self-reported background. Percentages in this table represent the number of fathers within each corresponding racial category of the father’s self-report and the mother’s report, over the column total. Numbers along the diagonal represent concordant reports. For example, 97.21% of fathers of newborns who self-report as White (non-Hispanic) are reported as White (non-Hispanic) by the mothers of those newborns, with similar levels of concordance among Black fathers (97.5%) and Hispanic fathers (93.6%). Discordance is relatively rare. In total, 187 fathers, or 5.4% of sampled fathers, report race/ethnic categories that do not match the mother’s report of their background. A little more than one percent of self-reported Black fathers are perceived as Hispanic by their newborn’s mother. Nearly two percent of self-reported Hispanic fathers are reported as White, while 3.53% are reported as Black. Less than 3% of all self-identified White fathers are reported as non-White by their newborn’s mother. Meanwhile, nearly 20% of Asian-identified fathers are reported using a non-Asian label by their child’s mother, and the majority of self-reported American Indian and Other Race fathers had discordant reports. The majority of mothers report these men as either White, Black, or Hispanic.

We next provide a more detailed exploration of the concordance of race and Hispanic origin reports (see Panel B), where we gauge race mismatch among fathers who are both self-reported and observed as Hispanic. Although there is still a substantial amount of concordance between the mother’s and father’s report of the father’s race, we see a relatively greater degree of racial mismatch among this group, with 307 Hispanic fathers perceived as a different racial category than they report. Around 64.95% of self-reported Hispanic-White fathers are reported as Hispanic-White by the mothers. We find even less concordance among Hispanic-Black fathers (56.67%). The highest degree of concordance occurs among self-identified Hispanic-Other Race fathers, as 71.09% are also reported as such by their newborn’s mother.

The forthcoming analyses define mismatch as any incidence of conflict between race and Hispanic origin responses. The prevalence of mismatch includes 187 cases from Panel A and 307 cases in Panel B, resulting in a total of 494 cases experienced by 14.25% of the sample. Our analyses capture this mismatch with a dichotomous measure indicating either discordance or matching between mothers’ and fathers’ reports of fathers’ race (1=mismatch and 0=matching reports).

Analytic Strategy

We begin examining the relationship between the independent variables and both types of racial mismatch by exploring the distributions of mismatch within each category of our independent variables. These patterns are then placed in a multivariate context, where we test the influence of each set of factors in shaping the likelihood of mismatch with logistic regression models, net of controls. The final model includes all independent measures. Firth logistic regression models are used to assess relationships. This modeling technique is ideal as our outcome is relatively unlikely among our respondents. Of the more than 3500 mothers, only close to 500 report any racial mismatch, with even smaller counts when divided by type (i.e. mismatches in which the mother perceives similarity or difference). Firth regression reduces issues of small-sample bias when conducting maximum likelihood estimation (Allison Reference Allison2012; Firth Reference Firth1993). Unfortunately, Firth regression is not possible with a trichotomous outcome, so the analysis is conducted with two logistic regressions that mirror a multinomial approach. Therefore, we predict all outcomes in separate analyses, restricting the analytic sample to matched reports and any mismatched reports (n=3467), matched reports and mismatched reports in which the mother perceives similarity (n=3316), and matched reports and perceptions of racial difference (n=3124), respectively. We have estimated comparable models with multinomial logistic analyses which produce highly similar results (analyses available upon request). All analyses were conducted with Stata 15.0.