Tipping the Multiracial Color‑Line: Racialized Preferences of Multiracial Online Daters

Abstract

Building on previous work on US multiraciality, we analyze the messaging patterns of Asian-white, Hispanic-white, and black-white multiracial heterosexual users on one of the largest mainstream dating websites in the USA. We consider how multiracials’ online dating behaviors reflect, accommodate or challenge racialized desirability hierarchies among heterosexual daters. The study’s results illustrate that Hispanic-white multiracial men show similar preferences to both their multiracial and monoracial in-groups, while Asian-white and black-white multiracial men most prefer their multiracial counterparts. Hispanic-white multiracial women, on the other hand, privilege whiteness and multiraciality, while Asian-white multiracial women show most preference for their multiracial in-groups. Overall, our findings illustrate that both multiracial men and women’s online dating behaviors illustrate a linked privileging of white multiraciality while they also reinforce a hierarchical ranking of racial desirability anchored by anti-Blackness.

Introduction

Though the USA has always been a nation of multiracial people due to long-standing histories of interracial mixing and the socially constructed nature of race itself, multiraciality has become especially significant in the twenty-first century. Interethnic marriage now accounts for one in twelve marriages in the USA, 2.9% of the American population reported two or more races on the 2010 US Census, and individuals who identify with two or more racial backgrounds will have the largest population increase in the coming decades (Colby and Ortman 2015; Pew Research 2015). Scholars speculate that this increase in multiracial identification will lead to an eventual break-down of racial boundaries, with greater numbers of mixed-race people willing to cross racial lines in the area of marriage, dating and childbearing (Alba and Nee 2003; Lee and Bean 2004; Littlejohn 2019; Qian and Lichter 2004, 2011).

Indeed, intimate desirability hierarchies around dating and marriage are one of the most long-standing racial divides in the USA (Davis 2010; Livingston and Brown 2017). Yet, scholarship has not adequately addressed whether increased rates of multiracial identification will translate to increased boundary-crossing in intimate life. Do multiracials see rigid divisions when it comes to whom they choose to date? Or are their choices more fluid than single-race individuals? Do they reinforce, accommodate or perhaps reject the pervasive bias and discrimination based on race and gender that has been documented in the literature on assortative mating (Curington et al. 2015; Feliciano et al. 2011; Lin and Lundquist 2013; Rafalow et al. 2017; Robnett and Feliciano 2011; Tsunokai et al. 2009)? Does this vary by gender or racial combination of specific multiracial background?

We explore these questions with data retrieved from one of the largest dating websites in the USA and analyze the online dating behaviors of black-white, Hispanic-white and Asian-white multiracial daters, three of the largest multiracial groups in the USA.¹ This approach allows us to examine not how multiracials are treated or what they say, but rather what they do. Our approach is novel with respect to existing studies on multiracials; most quantitative research focuses on the determinants of multiracial identity, but not on the agency of multiracial people themselves (see McGrath’s et al. 2016; Miyawaki 2015 for exceptions). Existing qualitative research, while equally informative, does not systematically compare across various multiracial groups (Buggs 2017, 2019; Khanna 2011; Newman 2017; Root 2004; Waring 2013). Given that Asian, Hispanics and blacks are subject to different processes of gendered racialization in the US, the question of how multiracial groups may differently negotiate a desirability hierarchy in dating provides important insight into how the strength of racial hierarchical boundaries may vary according to multiracial background.

Our findings suggest that black-white, Hispanic-white and Asian-white women and men tend to prefer their multiracial in-groups first and foremost. Except for in the case of Hispanic-white women daters, multiracial daters do not privilege whiteness over their own multiracial identity nor over non mixed-race minorities. Instead, we find that most multiracials are about equally interested in White daters as they are their corresponding minority counterpart. We also find that both multiracial men and women daters penalize a monoracial black status in online dating. These findings illustrate how multiracials co-construct a reconfigured racial desirability hierarchy that recognizes multiraciality as a category in its own right, yet one that is buttressed by the stigmatization of blackness.

Racialized And Gendered Desirability Hierarchies

Sociological research published within the past decade points to the persistence of gendered and racialized stratification in dating and marriage markets (see for example Lin and Lundquist 2013; Livingston and Brown 2017; Qian and Lichter 2011). Though most people date and marry within their own racial group, Asian and Hispanic women are more likely to pair with white heterosexual men than are black women. Meanwhile, white heterosexual women are more likely to pair with black and Hispanic men than Asian men (Bialik 2017). Racialized preference patterns also operate in the realm of online dating, one of the most popular venues for relationship formation today (Feliciano et al. 2009, 2011; Lin and Lundquist 2013; Rafalow et al. 2017; Tsunokai, Kposowa, and Adams 2009). White men and women are most favored by white and non-white daters alike, while black women and men are the only ones who consistently message one another (Lin and Lundquist 2013; Lundquist and Lin 2015; Rudder 2014). The messaging patterns in online dating are also gendered, with heterosexual men far more likely than women to initiate messages and white women exhibiting more rigid racial boundaries than white men (Hwang 2013; Lin and Lundquist 2013).

To explain these differences, gendered racial formation theory posits that intersections of race and gender shape societal ideals of desirability. Pervasive controlling images in the media stereotype black women as aggressive and unfeminine, while depicting black men as overly aggressive and hypermasculine (Collins 2004). The media has historically depicted Asian men as more feminine, while rendering Asian women as excessively submissive and feminine (Chou 2012). Meanwhile, these images celebrate whiteness as the epitome of idealized manhood and womanhood. Further, the intersection of these hegemonic notions of racialized gender frame the mixed-race body as “exotic” and attractive (Root 2004; Sims 2012; Waring 2013). Individuals from racial and ethnic marginalized groups may also deploy hierarchies of constructed desirability that resist, accommodate or internalize these desirability standards (Bany et al. 2014; Buggs 2017, 2019; Muro and Martinez 2018; Pyke 2010; Vasquez-Tokos 2017).

Where do multiracials fall within this hierarchy of desirability? Research on online dating finds that white/non-white multiracial daters receive a preference premium, with white and racial and ethnic minority monoracials more likely to reciprocate messages sent by Asian-white, Hispanic-white and black-white multiracial daters than by those sent by monoracial minorities (Curington et al. 2015). Yet, the question of how multiracial online daters themselves navigate this hierarchy has not been fully addressed. Qualitative research reveals that multiracial women and men’s preferences for racial and ethnic minority partners stems from cultural proximity and shared experiences, peer pressure, and the systematic rejection they encounter from white men and women throughout the life course (Khanna 2011; Korgen 1998; Rockquemore 2002; Song 2015). Other research finds that multiracial women daters reproduce racialized desirability hierarchies by privileging whiteness and drawing on anti-black ideology as they navigate intimate racial politics in dating (Buggs 2017, 2019). However, qualitative studies have yet to systematically compare the dating preferences of multiracial daters from varying racial and ethnic backgrounds.

The few existing quantitative studies on multiracial individuals’ dating and marriage habits glean important insight in this regard. Miyawaki’s (2015) analysis of the 2008–2012 American Community Survey data reports multiracials are predominately married to whites, with Asian-whites and American Indian-whites more likely than black-whites to be so. Littlejohn’s (2019) work finds that black multiracials are more likely to have a black partner than are their non-black multiracial counterparts. As much of this research analyzes marital and cohabitant patterns, less is known, however, about the initial processes of mate selection. Drawing from the existing theories and literature on multiraciality, we propose four hypotheses predicting how multiracial men and women may interact with daters on a mainstream dating website.

Minority Versus Whiteness Models

Historically, the legal codification of hypodescent, or the “one-drop rule,” accompanied anti-miscegenation laws in the USA and ensured that the offspring of enslaved black women would be born into chattel slavery (Khanna 2010). At its core, the rule of hypodescent provided a firm racial identity for African Americans of mixed racial background, prohibited mixed-race African Americans from marrying and partnering with whites, and further increased social distance between whites and blacks (Cox 1948; Davis 2010).

Some scholars suggest that the social consequence of hypodescent as it relates to social distance between whites and blacks still persists in a highly-racially stratified society like the USA (Gullickson and Morning 2011; Khanna 2010). Indeed, research reports that monoracials are most likely to perceive and categorize black-white multiracials as black (Chen and Hamilton 2012; Herman 2010; Ho et al. 2011, 2017; Qian 2004). Nikki Khanna (2011) similarly finds that the majority of her black-white interview respondents internalize a black identity, reasoning that whites largely treat and perceive anyone with black ancestry, regardless of phenotype, as Black. Survey data also reports that individuals of mixed white and black parental ancestry feel more socially aligned with their minority counterparts due to feelings of linked fate and shared marginalization, a reality that may impact their dating preferences (Davenport et al. 2016; Khanna 2010; Masuoka 2008; Pew Research 2015). For example, qualitative research finds that multiracial women with black ancestry express preference for black intimate partners and are likely to face rejection by white men in dating (Khanna 2011; Korgen 1998).

While multiracial daters may most align with their minority status, scholars argue that multiracials who are non-black are assimilating into the white mainstream, while, consistent with the "one-drop rule," black exceptionalism characterizes multiracials with black parental ancestry (Bonilla-Silva 2004; Bratter 2007; Gans 2012; Harris and Sim 2000; Lee and Bean 2007; Qian 2004; Qian and Litcher 2007). These arguments reason that non-black multiracials with a part white racial ancestry possess a symbolic racial or ethnic identity, while multiracials with a part black parental ancestry face greater social distance from whites due to their stigmatized racial identity. This may be because, historically, the legal assignation of a minority identity to mixed-race individuals was much more systematic for blacks, whereas it was situationally applied in the case of other ethnic and racial minority groups such as Asians and Hispanics (Gullickson and Morning 2011; Lee 2015). While racial identity construction is not the same as partner choice, groups who are racialized as closer to whites, may face fewer barriers to pairing with whites as they strive for a higher-status racial partner (Treitler 2013; Yancey and Yancey 1998).” Indeed, some quantitative research finds that non-black multiracials are more likely than multiracials with a part black racial identity to cohabit and marry whites (Littlejohn 2019; Miyawaki 2015).

A preference for whiteness in dating may also be gendered. The decision of whom to date and marry has historically been more consequential in terms of social status for women than for men (Coontz 2006). This is especially true for women of immigrant origin who may view intermarriage with whites as a strategy for assimilation (Vasquez-Tokos 2017, 2015). Hegemonic controlling images of idealized masculinity also often assign white men a higher desirability status, while men of color are often framed as devious rather than structurally diverted away from attaining resources associated with masculinity such as economic stability (Collins 2004). For example, Bugg’s (2019) recent work finds that some multiracial women’s preference for whiteness in online dating reflects their internalization of these hegemonic cultural notions which frame men of color, particularly black men, as undesirable partners. With respect to online dating, then, multiracial users may align with their minority counterparts and send or respond to their messages the most. We also propose the inverse of the Minority hypothesis. That is, multiracial daters—particularly non-black multiracial women—will be most open to white monoracial daters:

H1a

The Minority Model: multiracial daters will send and respond most frequently to their same-race monoracial minority counterparts.

H1b

The Whiteness Model: multiracial daters will send and respond most frequently to white monoracial daters.

Multiracial Model

Though some scholars speculate that black exceptionalism patterns exist among black multiracials, others argue that the social significance of hypodescent for part Black individuals has in fact waned in the post-civil rights era where an increase in interracial marriage, immigration and multiracial identification has led to a diversification of the color line (Bratter 2007; Gans 2012; Roth 2005). This is especially relevant given that scholars suggest that the US racial structure is changing to a more Latin American-like system where some groups—Asians, Hispanics and multiracials with a part white racial background—will be placed in a buffer group positioned above blacks, but beneath whites who hold the most privileged place in a status hierarchy (Bonilla-Silva 2004; Gans 2012). This argument calls for a proliferation of the racial structure in the USA, one that recognizes multiraciality as a different racial category in itself.

Indeed, research on the labeling decisions of multiracials report that blacks with mixed racial parental ancestry are increasingly veering away from a purely black identification and toward a multiracial label (Brunsma et al. 2013; Rockquemore and Brunsma 2002; Roth 2005). Given the absence of a legal system of hypodescent among non-black groups in the USA, mixed-race individuals with a more recent immigration history to the USA, such as Asians and Hispanics, may also embrace a multiracial identification over other racial options (Lee 2015). Increases in multiracial identification may also reflect increased group awareness of “being” multiracial in an increasingly diverse yet also racialized society like the USA (Waring 2017). In online dating where groups are exposed to a larger pool of potential daters, multiraciality may therefore be attractive to other multiracial daters given an assumption of shared experience and identity (Bernstein and De la Cruz 2009). And according to a survey conducted by Pew Research, multiracials feel a common bond with other multiracials and are more likely than monoracials to have multiracial partners (Pew 2015).

Another possibility, therefore, that has not been addressed in the literature is that multiracial daters will treat multiracials as in-groups and most prefer to date people who share their racial/ethnic combination. We therefore propose the following hypothesis

H2

The Multiracial Model: multiracial daters will send and respond most frequently to multiracial daters.

A strong version of this hypothesis is that multiracial daters would prefer other multiracial daters over single race white and racial minority daters even when they do not share the exact same racial backgrounds. A weak version is that multiracial daters only prefer those with the same racial backgrounds.

Bridging Model

An alternative possibility is that multiracial daters will be equally open to their monoracial and multiracials counterparts. This follows the popular racial “boundary blurring” argument predicting the blurring of social boundaries that often divide whites from non-Whites others due to increases in interracial intimate partnering and multiracial identification (Alba and Nee 2003; Gans 1999; Lee and Bean 2010; Qian and Lichter 2007). This theory reasons that interracial marriage and multiracial identification reflect a reduction in racialized social distance. Thus, mixed-race individuals may be more likely than monoracials to perceive porous boundaries among racial groups, a reality that may impact their intimate lives.

Past research provides some support. Multiracials express open attitudes towards interracial relationships (Bonam and Shih 2009; Bratter 2007; Qian and Litcher 2007; Roberts-Clarke et al. 2004). Research on the racial composition of Asian-white and black-white multiracial youth friendship networks also finds that multiracial adolescents are likely to bridge friendships among members of their single-race counterpart groups, displaying more inclusive behaviors than others (Doyle and Kao 2007). McGrath et al.’s (2016) analysis of multiracial online daters’ stated racial and ethnic preferences reports a bridging effect, with biracials generally more open to dating people “outside of their same racial/ethnic identity” (1936). In the context of online dating, this may play out with multiracials equally preferring both their white and minority counterparts due to a more general openness they display to monoracial groups with whom they share a racial and ethnic background. This logic, therefore, leads to the following hypothetical outcome:

H3

Bridging Model: multiracial daters will send and respond most frequently to whites, same-race minority monoracials and their same racial combination multiracial counterparts. The operation of the bridging model suggests multiracials treat all of these daters as in-groups.

Anti‑black Model

Finally, some scholars argue that the social recognition of multiraciality merely reinforces the pervasiveness of anti-blackness in contemporary society (Bonilla-Silva 2004; Gordon 1997; Newman 2017; Sexton 2008). This is because the distancing of blackness has historically underpinned a white and non-white binary, with non-white groups self-racializing as non-black, providing them relative privileges in a racial order where “Blackness is ‘the point from which the greatest distance must be forged’” (Gordon 1997, p. 53 cited in Kim 2018). Historically, this has played among racially marginalized communities, such as Mexicans, Chinese and European immigrant newcomers to the USA, who avoided partnering with blacks in order to attain a more privileged placement in the US racial hierarchy (Treilter 2013).

Past research on the online dating behaviors of various racial and ethnic groups also points to the durability of anti-blackness, with Asians and Hispanics ignoring black men and women who are typically the only ones who initiate contact with one another (Curington et al. 2015; Feliciano et al. 2009; Feliciano et al. 2011; Lin and Lundquist 2013; Lundquist and Lin 2015). In this research, we extend this proposition to multiracials; it is possible that they, too, follow anti-blackness in their online dating behaviors. This logic yields the following final hypothesis:

H4

Anti-blackness: multiracial daters will send and respond least frequently to monoracial blacks.

Method

Data and Sample

Our data are from one of the largest mainstream dating websites in the USA. Due to the data-use agreement with the company, we are unable to name the dating site in published work. Registered users create personal profiles, they can search and view other users’ profiles, and contact other daters through the site’s messaging system. Profiles contain basic information such as gender, sexual orientation, geographic location, age, race and ethnicity, height, body type, religion, language, lifestyle, and socioeconomic status. Unlike some mobile dating apps where interactions cannot occur until both daters have swiped right on one another, indicating they are already interested in the other, our data is able to measure initial interest of individual daters indicated by a first message. Moreover, the criteria used to select which profiles to display are age, geography, sexual orientation, and a matching score derived from optional personality questions. That said, the users are still able to search other users by ethnic or racial status based on their own preferences.

We identify the users’ own racial identity based on information from their personal profiles. They can check off any or all of the following ten “ethnicity” boxes: Asian, Middle Eastern, black, Native American, Indian, Pacific Islander, Hispanic, white, Other, and Undeclared. We define individuals checking off more than one box as multiracial. For example, we categorize users who checked Hispanic and white and nothing else as Hispanic-white. Similarly, users who checked Asian and white are categorized as Asian-white, and so on.² We dropped 24% of the original sample from the analysis with blank racial categories; analysis of these cases show that most were unfinished profiles of inactive daters. About 4% of the users identified themselves as black in contrast to 12% in the U.S. population. About 5% of the users identified as Hispanic in contrast to 16% of the U.S. population. Asian, white, and multiracial daters, on the other hand, are over-represented. This pattern is likely driven by the website’s appeal to young professionals.

The original dataset consists of around 200-million messages exchanged among 9 million registered users from November 2003 to October 2010. Our analysis focuses on messages sent by heterosexual Asian, Hispanic, black, white, Asian-white, black-white and Hispanic-white women and men and the responses they received from heterosexual biracial groups (Asian-whites, black-whites, and Hispanicwhites). We focus on multiracial with a white and racial/ethnic minority background for two reasons. For one, they are the vast majority of multiracials in our sample, as well as in the US population as a whole. Over 80% of multiracial Americans report white as part of their ethnic composition and 93% report being biracial (CensusScope 2014). Secondly, we restrict our analysis to heterosexual white/non-white multiracials for practical reasons of sample size. While there are other multiracial daters (non-white and/or non-heterosexual) on the website, their numbers are small and we lack sufficient interaction between them and others for complex modeling.³

Our analysis sample consists of 6,919,625 initial messages and responses exchanged in the largest metropolitan statistical areas of the USA.⁴ Census data from 2010 indicate that multiple-race reporting is least concentrated outside of metropolitan counties, and multiracial house-holds are predominately located in racially diverse census tracts in major U.S. metropolitan cities (U.S. Census Brief 2010). For this reason, we limit our analysis to the 20 largest U.S. cities. In terms of representativeness and generalizability, our data is truly representative of the self-selected groups of all users who utilize the website; however, our data do not allow us to make gross generalizations to society at large. Still, traditional quantitative studies which focus on multiracial daters’ stated preferences in online dating are unable to observe the actual interactions and behaviors of multiracials (McGrath’s et al. 2016). Our work, on the other hand, provides important insight into the direct ways in which multiracials interact with others.

This article focuses on men’s messaging behaviors and women’s responding behaviors, since this is the predominant form of interaction among heterosexual daters (Lin and Lundquist 2013).⁵ The reasoning is this: Because it is more normative for heterosexual men to send first messages, we will get a better sense for their true preferences than when they are limited to responding to the smaller pool of the unusual women who messaged them first. The reverse reasoning applies to women. Heterosexual women who take the initiative to send a first message are rare compared to women who wait to respond. While we are not certain that a response represents a genuine interest, only 8.5 percent of the initial messages had a response, which indicates responses are socially meaningful. The widely adopted norms of online dating overwhelmingly point to nonresponse as a robust measure of rejection (Lin and Lundquist 2013). To control for cases when a woman dater has only a small pool of daters to respond to, we use a popularity measure, the average number of messages received.

Table 1 presents the number of observations and the likelihood of messaging and receiving a response by gender and by racial identity. Similar to previous findings, it shows that more initial messages are sent from men to women, and men are more likely than women to respond to an initial message. Table 1 also provides preliminary insight into how multiracial daters respond to other daters. Multiracial women in general are more likely to respond to white men than minority men that share their racial status. In addition, they also demonstrate some preferences for other multiracial men even when the racial composition is different from their own. Multiracial Asian-white, black-white and Hispanic-white men in general show preferences for mono- and biracial women who share their racial background. However, a level of anti-blackness is observed across the three groups. Black-white men are more likely to send messages to Asian and Hispanic women than to their black counterparts.

Table 1.

Number of observations and the likelihood of messaging (in parentheses) by gender and ethnic-racial identity

	Messenger’s ethnic-racial identity
	Asian-White	Black-White	Hispanic-White
To	Panel A: men messaging to women
Asian	575,468 (0.0012)	226,887 (0.0010)	1,148,333 (0.0007)
Black	536,563 (0.0003)	241,762 (0.0008)	1,261,878 (0.0003)
Hispanic	541,218 (0.0007)	227,819 (0.0015)	1,384,886 (0.0014)
White	743,909 (0.0012)	332,394 (0.0013)	1,775,614 (0.0014)
Asian-White	205,909 (0.0014)	79,106 (0.0010)	406,432 (0.0008)
Black-White	105,301 (0.0006)	47,198 (0.0022)	225,288 (0.0007)
Hispanic-White	419,271 (0.0009)	174,559 (0.0013)	949,832 (0.0013)
To	Panel B: women responding to men
Asian	548 (0.0748)	3660 (0.0388)	3341 (0.0338)
Black	2761 (0.0594)	1690 (0.0331)	4680 (0.0387)
Hispanic	1252 (0.0639)	2723 (0.0389)	11,066 (0.0447)
White	14,399 (0.0876)	33,416 (0.0654)	80,266 (0.0671)
Asian-White	250 (0.1000)	220 (0.0636)	545 (0.0514)
Black-White	143 (0.0909)	737 (0.0516)	848 (0.0649)
Hispanic-White	452 (0.0686)	1036 (0.0492)	3900 (0.0664)

Controls

We control for the following confounding factors: demographics, lifestyle, and online activity. The descriptive statistics for our sample of daters, broken down by gender and racial composition, are shown in Tables S1A through S1D in the online supplement. Demographic controls include age, height, body type, and level of education. On average, daters are between 23 and 28 years old. Women tend to be more represented in the youngest age bracket (< 23). Daters who self-identify as multiracial are younger, on average, than monoracials. Black, white, and black-white multiracial men and women report the tallest heights. Many daters, between 30 and 40%, do not report body type. Among those who do, there are notable gender differences. Except for Asian and Asian multiracial women, women are more likely than men to report their body type as over-weight. All men, however, are most likely to report themselves as fit. Education levels vary more by race than by gender. Asian men and women daters have the highest levels of education; Hispanics the lowest. Multiracial daters have slightly lower education levels, which is likely the result of younger age.

Lifestyle controls include smoking, drinking, drug use, and parental status/preference. Online activity controls include how many personality questions the sender answered and how many photos the sender uploaded. The expectation is that an initial message is more likely to garner a response if the profile is more developed. Hispanic women and black men tend to answer the least personality questions; Asian-white men and women tend to answer the most. Multiracial daters of both genders upload the most pictures of themselves; monoracial Asian and black daters upload the fewest. We also control for the number of messages the receiver was sent per 100 days, and the matching score between the sender and receiver. The former is expected to have a negative effect; the latter should have a positive effect on the outcome of interest.

Another potential confounder is the attractiveness of daters, about which we lack reliable data.⁶ However, for this to be a confounder, there needs to be substantial racial variance in “objective” attractiveness. We believe it is more appropriate to treat attractiveness not as a confounder but as a consequence of racial status. In other words, attractiveness is not an independent variable but itself a product of racial hierarchy. In a separate analysis we focus on daters without photos and find that the racial hierarchy remains similar (Curington et al. 2015).

Analytic Approach

Our main analysis uses Generalized Estimating Equations (GEE) to predict the likelihood of sending and response after controlling for the interaction of each party’s attributes. By adding these controls, we can more precisely identify how multiracial daters interact with daters of various racial backgrounds differently net of observable factors. We estimate a separate model for each race-gender group such as Asian-white multiracial women and Asian-white multiracial men, etc., and thus allow racial preference to vary by racial identity and gender. We specify the model predicting the sending behavior for users of group k as

$$ln\left[\frac{P\left(y_{i,j}=1\right)}{1-P\left(y_{i,j}=1\right)}\right]=a_i+\sum _k^K{\beta }_{1,k}{R}_k+\sum _q^Q{\beta }_{2,q}{X}_{q,i,j}+\sum _r^R{\beta }_{3,r}{X}_{r,i}+{\beta }_4{X}_{4,i}+{\beta }_5{W}_{i,j}+{\epsilon }_{i,j}$$

where $y_{i,j}$ denotes whether an initial message was sent from $i$ to $j$, $R$ denotes a series of indices for the 49 racial combinations of the sender and the receiver, $X_{q,i,j}$ denotes the combination demographic and lifestyle attributes of users $i$ and $j$, $X_{r,i}$ denotes potential receiver $j$’s number of questions answered, photos uploaded, and time spent on site, $X_{4,j}$ denotes the number of messages sender $i$ sent per 100 days, and $W_{i,j}$ denotes the length of overlap between users $i$ and $j$ on the website. We are also interested in how the likelihood of a response varies by the racial and gender identity of the sender and that of the receiver. We specify the models as follows:

$$\ln \left[\frac{P\left(y_{i,j}=1\right)}{1-P\left(y_{i,j}=1\right)}\right]=a_i+\sum _k^K{\beta }_{1,k}{R}_k+\sum _q^Q{\beta }_{2,q}{X}_{q,i,j}+\sum _s^S{\beta }_{3,s}{X}_{s,i}+{\beta }_4{X}_{4,j}+{\beta }_5{M}_{i,j}+{\epsilon }_{i,j}$$

where $y_{i,j}$ denotes whether user $j$ responded to the initial message sent by user $i$, $R$ denotes a series of indices for the 49 racial combinations of the sender and the receiver, $X_{q,i,j}$ denotes the combination demographic and lifestyle attributes of users $i$ and $j$, $X_{s,i}$ denotes sender $i$’s number of questions answered and photos uploaded, $X_{4,j}$ denotes the number of messages responder $j$ received per 100 days, and $M_{i,j}$ denotes the matching score determined by the personality questions.

Because men and women select potential mates on different criteria, we estimate two separate models for women and for men. Moreover, because decisions to send or respond to a message are nested within individual $j$, we estimate the outcome by fitting a series of generalized estimating equations (GEE) (Hanley et al. 2003; Liang and Zeger 1986; Zuur 2009) with an exchangeable correlation structure. This approach analyzes dyad dependency, taking into account the characteristics of millions of daters within the same city who may or may not make contact with one another.⁷ Specifically, we model two forms of interactions: the likelihood of messaging and the likelihood of responding after controlling for an array of attributes of both members of the dyad. The odds ratios at the mean of the covariates are generated from these models to compare preferences among different racial groups. In the following section, we present the odds ratios of messaging with confidence intervals calculated with the Delta method.

Results

Figure 1 presents the odds ratio of Asian-white men messaging women and Asian-white women’s odds ratio of responding to messages sent by men. The dark gray dots represent the estimates while the intersecting gray lines depict 95% confidence intervals. Points that fall greater than one indicate a higher likelihood of messaging or responding to Asian, black, Hispanic, white and multiracial daters than to their own counterparts (Asian-whites), after accounting for both his and her profile and personality matching factors. Points lower than one indicate lesser interest compared to their counterparts. Where confidence intervals do not cross the odds ratio of 1.0 level, the tendency toward messaging versus not messaging one’s in-group is statistically significant.

Fig. 1.

Asian-White men’s messaging and women’s response

Looking first at Asian-white men’s patterns of messaging on the left hand side of Fig. 1 and taking into account the overlapping confidence intervals, Asian-white men send the most messages to Asian-white women. They also send second most messages to Asian women and the least messages to black women. Asian-white men’s odds ratio of sending to black women is approximately 0.1 relative to sending to Asian-white women, which is also substantially smaller than the odds ratio for sending to Asians, which is around 0.8. This messaging behavior best fits with the Multiracial and Anti-Black Model, followed by a tendency towards the Minority Model, or Hypothesis 2, 4 and 1a.

We now turn to the odds ratio of response for Asian-white women, as displayed on the right side of Fig. 1. Due to the large confidence intervals, it is less clear which sorting mechanism determines their responsiveness. Based on point estimates alone, Asian women appear, like their counterparts, to respond first and foremost to other Asian-white multiracial daters. However, the overlapping confidence intervals at the 95% level suggest that this is not statistically different from their propensity to also respond to whites, Asians and black-white men. They also appear to respond less frequently to messages sent by Hispanic, Hispanic-white and black men. For example, their odds ratio of responding to blacks, Hispanic and Hispanic-whites relative to responding to Asian-whites is approximately between 0.6 and 0.7. Generally, we observe a weaker version of the Multiracial Model in operation (Hypothesis 2), with a tendency towards the bridging Model (Hypothesis 3).

We speculated that whiteness may be more desirable to non-black multiracials and the Minority model, or Hypothesis 1a, would apply most to black-white multiracials’ online dating behaviors. Now we consider whether this caveat plays out. We present the same odds ratios for black-white men and women in Fig. 2. The same logic applies here; where confidence intervals do not cross the 1.0 level, the effects are statistically significant. Beginning with men’s messaging patterns displayed on the left side of the figure, black-white multiracial men are most inclusive of their black-white counterparts and send less frequently to all other women daters. This is the strongest in-group preference we have seen thus far. Like Asian-white men, however, they also appear to send the least messages to black monoracial women. For example, black-white men’s odds ratio of sending to blacks relative to sending to black-white multiracials is 0.3, which is smaller than the odds ratio for sending to white women (around 0.5). However, the odds ratio of sending to a non-black multiracial women never increases above 0.6. In this sense, we again observe the Multiracial and anti-Black Model at play (Hypothesis 2 and 4).

Fig. 2.

Black-White men’s messaging and women’s response

Figure 2 also illustrates the odds ratio of black-white women’s responsiveness to men’s messages. This figure shows that the odds ratio of responding to Asian-whites and white men relative to black-white men is around 1.5. However, as the confidence intervals cross the 1.0 level, the difference is not statistically significant at the 95% confidence level. Given the large error for these estimates, we cannot discern that black-white women have any clear racial preferences and are therefore unable to associate any of our hypotheses with their results.

Figure 3 lastly presents the odds ratio of messaging and response for Hispanic-white men and Hispanic-white women. Again, probabilities further to the right in each figure indicate that they are more preferred to groups with smaller probabilities to the left. Beginning with the odds ratio of men’s response (left panel), here we see the Bridging Model in operation; Hispanic-white multiracial men send messages to Hispanic, white and Hispanic-white women at a relatively equal rate. Despite this inclusive patterning of behavior, however, they do appear to send relatively fewer messages to black women when compared to all other women. For example, Hispanic-white men’s odds ratio of sending to black women relative to sending to Hispanic-white multiracial women is 0.2. In this sense, we again confirm the Anti-Black hypothesis, or H4.

Fig. 3.

Hispanic-White men’s messaging and women’s response

Hispanic-white multiracial women’s patterns of response depart from that of Hispanic-white multiracial men. As illustrated in the right panel of Fig. 3, the dividing line vacillates across white and multiracial status (Whiteness Model and Multiracial Model), as they tend towards a preference for white men, followed by all other multiracial daters. While this is the only clear instance of a generic multiracial preference thus far, their treatment of Asian-white and black-White men is not significant at the 95% confidence level. On the other hand, they do appear to respond less frequently to Asian, black and Hispanic monoracial men, though they appear to respond to black monoracial men the least. For example, Hispanic-white women’s odds ratio of responding to black men relative to responding to Hispanic-white multiracial men is around 0.6, which is smaller than their odds ratio of responding to Hispanic men (0.8) and Asian men (0.7). Overall, we find general support for the Multiracial and Anti-Black Models; Asian-white, Hispanic-white and black-white multiracial men and women are generally most open to other multiracial online daters and exclude monoracial blacks in online dating. Meanwhile, support for the white, Minority, and Bridging Models depends on the specific combination of gender and multiracial racial group. In the following, we theorize the driving force behind these observed patterns in online dating (see Table 2 for a summary of the results).⁸

Table 2.

Hypotheses confirmation, by group

Group	Minority	Whiteness	Multiracial	Bridging	Anti-Black
Status	H1a	H2a	H2	H3	H4
Men  Asian-White	✓		✓		✓
Black-White			✓		✓
Hispanic-White				✓	✓
Women  Asian-White			✓	✓*	✓
Black-White		✓*	✓		✓*
Hispanic-White		✓	✓		✓

^*Not statistically significant at the 95% level

Discussion

The Multiracial Model best characterizes Asian-white men’s messaging patterns as they are most inclusive of Asian-white multiracial women. Asian-white men also display a weaker version of the Minority Model, as they send almost as many messages to Asian monoracial women. One interpretation of these findings is that Asian-white men perceive more rigid boundaries between them and white women, a speculation supported by the research documenting the penalization monoracial Asian men face by white women in online dating (Feliciano et al. 2009; Rudder 2014). In this sense, Asian-white men may message white women less than they do Asian and Asian-white women due to anticipatory rejection. Black-white men’s messaging behavior also displays the Multiracial effect; they, foremost, message black-white multiracial women and message other multiracial and monoracial women less frequently. Together, these findings suggest an emerging group solidarity among some multiracial daters.

Hispanic-white multiracial men’s messaging behavior is also noteworthy given that they are equally open to Hispanic, white and Hispanic-white women. This operation of the Bridging effect suggests that Hispanic-white men may not perceive significant social distance between themselves and Hispanics or whites. This finding also poses questions concerning whether or not Hispanic-whites see themselves as having a symbolic racial or ethnic identity (Lee and Bean 2004). Despite this variation across men’s behavior, neither multiracial male group exclusively privileges whiteness over all other identities, while black women appear to be the most excluded by all multiracial men. Additionally, Asian-white and Hispanic-white men are relatively less likely to send messages to black-white multiracial women than to other multiracial women. In this sense, we document multiracial men’s reinforcement of gendered Anti-blackness in online dating.

Unlike Asian-white and black-white multiracial men who illustrate disinterest in whiteness and Hispanic-white men who treat Hispanic, Hispanic-white and white women similarly, Hispanic-white multiracial women tend toward a white preference in online dating. This may be because the decision of whom to date and marry has historically been more consequential in terms of social status for women than it has been for men (Spickard 1991). Research also finds that Hispanic women face more gendered policing by family members to pair with partners of lighter complexions than do Hispanic men (Stephens et al. 2013; Vasquez-Tokos 2015; Vasquez-Tokos 2017). Hispanic women are also likely to face familial disapproval and sanctioning when they pair with African American men (Bany et. al. 2014; Garcia et al. 2012). Further, Hispanic women may gravitate towards white men in dating or marriage because they assume that white men are gender progressive and that Hispanic men are gender conservative (Muro and Martinez 2018; Vasquez-Tokos 2017). In this sense, Hispanic-white women’s greater responsiveness to white men may also be shaped by these racialized and gendered dynamics as well.

Further, though we find that all multiracial women generally penalize black monoracial men in online dating, they do not display this exclusionary behavior in their treatment of white-black multiracial men. This finding is especially significant given that much of the focus on multiracial dating has been on the exotification of multiracial women (Buggs 2017; Root 2004). Yet, emergent research does find that black multiracial men are also perceived and constructed by others as desirable (Newman 2017).

Additionally, unlike Hispanic-white women, Asian-white women do not privilege whiteness and, instead, are most responsive to their Asian-white counterparts. This is significant given that past research found that white men are especially inclusive of Asian-white women, responding to their messages the most (Curington et al. 2015). Additional research on online dating also finds that monoracial Asian women show most interest in white men (Lin and Lundquist 2013). So what explains Asian-white women’s preference for Asian-white multiracials? For one, though past research suggests that Asian-white women are desired by white men in online dating, Asian and multiracial desirability, as we have argued, may be couched in ethnic and racial fetishization (Cho 1997; Chou 2012; Root 2004; Waring 2013). Asian-white multiracial women may simply seek to make unproblematic connections that are free of sexual-racial fetishization. Thus, Asian-white women’s preference for Asian-white men may be driven by the intersection of these gendered dynamics with an emerging group solidarity among Asian multiracials that we observe in other multiracial groups’ online dating behavior as well.

Indeed, Asian-white men and black-white men most prefer their multiracial in-groups, and multiracial women generally privileged multiraciality. To identify as mixed race and to most prefer mixed race suggests the possibility that multiracials differentiate themselves and other mixed-race users from their minority monoracial counterparts. We see this behavior as both an inclusive and exclusive behavior; it is inclusive because it suggests that multiracials align with other multiracial daters, perhaps a reflection of greater public recognition and validation of multiracial identities in the contemporary cultural landscape (DaCosta 2007; Sexton 2008). Multiracials may also prefer other multiracial daters due to the shared experience of occupying a liminal space in a US racial schema which often rests on racial dichotomy, despite the greater social representation and recognition of mixed-race people (Bernstein and De la Cruz 2009). However, as we will discuss below, we also see this as exclusive behavior given that it occurs alongside multiracial online daters’ exclusion of monoracial blackness.

Conclusion

As scholars predict that individuals who identify with two or more racial backgrounds will have the largest population increase by the year 2060 (Colby and Ortman 2015), one of the key questions is whether greater social recognition of multiraciality represents a transcendence of racial hierarchies, or whether it reinforces or alters them in some way. Our data suggest that most multiracials tend to prefer their own multiracial in-groups first and foremost, which suggests that they have developed a distinctive multiracial identity. Despite past work finding part white multiracials to be privileged by whites in online dating, except for in the case of Hispanic-white women, there does not appear to be a reciprocated privileging of whiteness over their own multiracial identity, nor over their monoracial minority identity (Curington et al. 2015). We instead find that most daters are about equally interested in monoracial white daters as they are their corresponding racial minority counterpart. This could be read as a resistance to long-standing hegemonic notions of desirability that privilege white masculinity and femininity.

However, one way in which multiracial daters internalize larger racial hierarchies is in their treatment of black daters. In almost every case, black monoracials are the most excluded group by multiracial daters. These findings line up with discussions predicting a shifting racial order, one towards a tri-racial system where multiracials are positioned as an interstitial category in the U.S. status hierarchy (Bonilla-Silva 2004). Bonilla-Silva (2004) purports that some multiracials (particularly those with a part white parental background) are “honorary whites” and will actively distance themselves from collective blacks in a racial order that confers varying degrees of privilege to those who tilt closer to whiteness. Newman (2017) extends this argument and purports that the “insertion of multiraciality into the racial order reconfigures that structure” (16) in such a way that a racial hierarchy anchored by anti-blackness is reinforced, while previously stigmatized groups, such as black multiracial men, enjoy a more privileged position in a status hierarchy. In this sense, the operation of the Multiracial Model may indicate the affirmation of a multiracial identity, yet, as we have found, it does not fully translate to a breaking down of a racial desirability hierarchy that has long been anchored by black stigmatization. Both Multiracial and Anti-Blackness Models indeed operate simultaneously.

This study’s limitations suggest various avenues for fruitful research on multiraciality. As the current study draws from quantitative interactional data, future qualitative research is needed to assess multiracial men’s and women’s own interpretations and negotiations of the behavioral patterns we document (Buggs 2017, 2019; Curington 2020; Waring 2013). Second, this study focused only on individuals who identified as white and some other race and therefore does not capture the behavior of non-white multiracials. Though the majority of self-identified multiracial individuals identify as partly white in the USA, it would be useful to know how non-white multiracial behavior departs from that of white multiracials. Recent research focused on preferences does provide some insight (e.g. McGrath et al. 2016), but, as we have argued, preferences may not reveal actual behavior. Another question is whether queer multiracial daters also reproduce racialized desirability hierarchies in their treatment of daters. Furthermore, photographic data would also tell us how phenotype shapes these preferences. Finally, as online dating sites ask questions regarding self-identification, as opposed to an individual’s racial ancestry, our analyses do not consider individuals of mixed racial parentage who self-identify with one race.

Despite these limitations, our findings reveal that the behaviors of self-identified black-white, Asian-white and Hispanic-white heterosexual daters do not push back against the black devaluation that has been well documented in the literature on online dating (Feliciano et al. 2009; Lin and Lundquist 2013; Lundquist and Lin 2015; Robnett and Feliciano 2011; Rudder 2014). Instead, their behaviors privilege multiraciality while reinforcing a hierarchical ranking of racial desirability, one anchored by anti-monoracial blackness.