Data Considerations

Although internet use has expanded exponentially in recent years, internet users are still a select sample, and this is especially true among Blacks and Latinos (Jayajit and Bosman, Reference Jayajit and Martin Bosman2005; Mack Reference Mack2001). For example, around the time of our data collection, 71% of Whites used the internet, compared to only 60% of Blacks (Fox and Livingston, Reference Fox and Livingston2007). Internet use among Latinos varies by language: English-dominant or bilingual Latinos use the internet at rates similar to Whites, but only 32% of Spanish-dominant Latinos use the internet (Fox and Livingston, Reference Fox and Livingston2007). However, the most inequality in internet use is found by socioeconomic status; those with the lowest income and education levels are far less likely to go online across and within all racial/ethnic groups (Fox and Livingston, Reference Fox and Livingston2007; Martin and Robinson, Reference Martin and Robinson2007). Thus, our data cannot be generalized beyond the population of U.S. internet users in these four metropolitan areas, who have higher socioeconomic status than the general U.S. population.^{Footnote 9} The sample of Blacks and Latinos is even more selective than the sample of Whites and Asians in this regard, and the sample of Latinos also underrepresents recent immigrants.

In addition, although internet dating has become, by most accounts, a mainstream practice in recent years (Rosenfeld and Thomas, Reference Rosenfeld and Thomas2012; Sautter et al., Reference Sautter, Tippet and Philip Morgan2010), internet daters may still be a select group of single internet users. However, recent survey research suggests that internet daters do not differ in socioeconomic or demographic characteristics (such as gender, race, or education) from single internet users who do not use internet dating services (Sautter et al., Reference Sautter, Tippet and Philip Morgan2010). The strongest determinants of internet dating among single internet users were whether respondents were actively looking for a partner and whether they knew someone who had tried online dating (Sautter et al., Reference Sautter, Tippet and Philip Morgan2010). Thus, our data are fairly representative of single internet users in these U.S. metropolitan areas, who tend to be more educated and skilled in writing in English.

A further sample selection issue was the possibility of self-selecting minorities who are especially open to interracial dating. This might be the case if Yahoo Personals were dominated by Whites; minority daters also have the option to use ethnic-specific websites. However, Whites were actually underrepresented on the website compared to their representation among internet users in all the regions except for Los Angeles (where their representation approximates the percentage of internet users).^{Footnote 10} In general, the racial distribution of internet daters on Yahoo Personals closely approximates the racial distribution of internet users in the four metropolitan areas (table available upon request). Thus, we have little reason to expect that the racial makeup of the Yahoo dating pool would affect racial preferences any more than racial preferences are shaped by the racial makeup of the communities in which the daters live, a factor we include in our analysis.

We also consider the possibility that some internet daters might be especially choosy about who they date (and thus unable to find their preferred dates off-line). We account for this by examining differences in racial exclusion and inclusion controlling for how choosy the dater is in general, which is measured by the percentage of characteristics for which the dater states a preference.

Unfortunately, the website does not distinguish between those who are looking for a serious relationship, versus those who are seeking only a casual relationship. Given prior research, which shows that interracial relationships are less likely than same-race relationships to lead to marriage (Joyner and Kao, Reference Joyner and Kao2005), our results do not necessarily represent willingness to engage in serious interracial relationships. However, willingness to even casually date someone of another racial group indicates a certain level of social acceptance and is necessary for a more serious relationship to develop.

We were also concerned that there might be some misrepresentation in daters’ stated preferences. However, if online daters do misrepresent racial preferences, they are likely to do so in the direction of including racial groups whom, in reality, they are not open to dating. Prior research examining actual online contact on a dating Website shows that daters who do not state any racial preferences in their profiles nevertheless tend to discriminate against members of different racial groups (Hitsch et al., Reference Hitsch, Hortaçsu and Ariely2010b). Also, preferring “any” racial group is the default on the website; daters who are in a rush to complete a profile may choose to disregard the choices. Others may wish to appear politically correct. For these reasons, we make an analytical distinction between those who state they are open to dating “any” racial group and those who stated particular choices and excluded and included particular groups. Given that these are real individuals searching for a date, daters would have no reason to exclude groups that they are actually open to dating, and thus we are confident that the patterns we find accurately represent true preferences, even if they underestimate the degree to which particular racial groups are actually excluded.

Despite their limitations, the data hold a number of advantages. Even if not representative of the general population, internet dating pools are not trivial: approximately sixteen million Americans use such services; nearly three-quarters of all internet users who are single and looking for romantic partners have used the internet to find dates and Yahoo Personals had more than six million unique visitors each month (Madden and Lenhart, Reference Madden and Lenhart2006). Morever, the internet is now one of the primary ways that couples meet (Rosenfeld and Thomas, Reference Rosenfeld and Thomas2012). There is also remarkable consistency between patterns of racial preferences stated in internet profiles and patterns of interracial unions in marriages and cohabitating relationships (Hitsch et al., Reference Hitsch, Hortaçsu and Ariely2010a). This suggests that preferences stated on the internet correspond to preferences that drive union formation regardless of where couples meet. Most importantly, these data provide a rare opportunity to reveal preferences as stated in a real-life situation, unlike attitudinal surveys or social distance scales based on hypothetical scenarios.

Measurement of Perceived Race

Our key independent variable is the perceived race of the Latino daters. Because the way outsiders categorize others by race may vary by the observer’s own characteristics, especially their own racial background (Harris Reference Harris2002), three research assistants from different racial backgrounds independently assessed the daters’ race based on their photographs.^{Footnote 11} The coders were provided copies of the self-identified Latino daters’ profiles and were instructed to code the racial category that best described the dater, based solely on his/her appearance in the photograph. The racial categories the coders could choose from were White/Caucasian, Black/African American, Asian, Latino/Hispanic, or Other. Although there is debate about whether Latinos are considered a racial or ethnic category, we followed Wendy Roth (Reference Roth2010) in referring to Latinos as a racial category because individuals recognize a phenotypic “Hispanic” racial type (characterized by brown skin and a mix of European, indigenous, and/or African features). Experimental research has confirmed that there is a prototypical “Hispanic” or “Latino” appearance (Wilkins et al., Reference Wilkins, Kaiser and Rieck2010). None of the coders expressed confusion or questioned whether Latino/Hispanic was a valid racial category.

The coders were not told that these daters self-identified as Latino and were explicitly instructed not to look at any parts of the profile except for the photograph.^{Footnote 12} We note that while many previous studies only examine phenotypic dimensions of race (i.e., skin color), the observers here may have relied on self-presentations of cultural identities as well (as expressed through clothing, hairstyle, etc…). Thus, our study captures how observers categorize others based on both physical and cultural aspects of appearance. Racial classifications may also be affected by social status cues. For example, previous research has shown that observers are more likely to categorize individuals as White if they are dressed in a business suit, and Black if dressed in a janitor’s uniform (Freeman et al., Reference Freeman, Penner, Saperstein, Scheutz and Ambady2011). Thus, the racial categorizations here may also be based on implicit associations between social status and race to the extent that status is indicated in a photograph.

Agreement between any two coders ranged from 68–74%.^{Footnote 13} In cases of disagreement, if two coders agreed, we used that racial category as the observed race. If all three raters disagreed (4.5% of all cases), we recoded the dater’s observed race into a residual “ambiguous” category, which was then collapsed into the “other” category. Figure 1 shows the coders’ assessments of the observed race of the self-identified Latino sample. We see that about 72% of the Latinos were observed as “Latino” in appearance, while 14% were perceived as “White,” 7% as “Black,” and 7% as Other/ambiguous.

Fig. 1. Observed Race of Self-Identified Latinos, Yahoo Internet Daters, N=1528

Fig. 2. Predicted Probabilities of Including Whites and Blacks as Possible Dates, by Self-Identified Race and, for Latinos, Observed Race, Yahoo Internet Daters

Notes: LL=significance test relative to observed Latinos, W=significance test relative to self-id Whites, B=significance test relative to self-id Blacks, ^nsp>.10, ⁺p < .10, *p < .05, **p < .01, ***p < .001.

Multinomial regression models control for gender, age, metropolitan area, racial composition of municipality, education, body type, religion, Spanish language, choosiness, preferences for religion, body type, education, and height.

Control Variables

Because previous research has indicated that personal characteristics shape racial preferences (Feliciano et al., Reference Feliciano, Lee and Robnett2011; Feliciano et al., Reference Feliciano, Robnett and Komaie2009; Wilson et al., Reference Wilson, McIntosh and Insana2007; Yancey Reference Yancey2007, Reference Yancey2009), we controlled for the dater’s gender, age (daters average around thirty-three years old), education (coded as high school graduate or less, some college, college graduate, or post-graduate), political views (coded as liberal or very liberal vs. middle of the road, conservative or not political), religion (Christian, not religious, other, or no answer), Spanish language knowledge (speaks Spanish, does not speak Spanish, and no answer), and body type (slim/slender/average, thick/a few extra pounds/voluptuous, and no answer).

We also wanted to account for the possibility that daters were using race as a proxy for their preferences for other characteristics, such as body type or education. For example, some daters might exclude Blacks not because they care about race per se, but because they want to date someone highly educated, and therefore engaged in statistical discrimination based on the knowledge that Blacks, on average, have lower levels of education. To account for this, we control for whether the dater stated a preference for educational attainment, religion, body type, or language.^{Footnote 14} We also considered that daters expressing racial preferences were just more particular in general about their date’s characteristics by including a control variable for how selective the dater was about their date’s characteristics overall, based on the percentage of the nineteen other items for which the dater expressed a preference.

We also considered that racial preferences might just result from exposure to different racial groups in their communities. To address this, we examined the racial composition of the community that each dater reported as their residence on their profiles. Using data from the 2005 American Community Survey (U.S. Bureau of the Census 2005), we collected data on the percent non-Hispanic White, percent non-Hispanic Black, and percent Latino in each town/municipality.^{Footnote 15}

Dependent Variables

We coded racial preferences into three sets of multinomial dependent variables. Daters chose more than four hundred unique combinations of racial groups as preferred dates. For example, daters might indicate that their “match” should be Caucasion (White), Asian, or Hispanic/Latino, others might indicate that their “match” should be African/American (Black), Inter-racial, or Other, and others might indicate that their “match” should only be Hispanic/Latino. To simplify, we focus on three multinomial outcomes. First, we examine a preference for racial homophily using a four category dependent variable: (1) includes own racial group and other groups; (2) includes own racial group only; (3) excludes own racial group/includes only other racial groups; and (4) no stated preference. Second, we examine acceptance of Whites using a three category outcome: (1) excludes Whites, (2) includes Whites, and (3) no stated preference. Third, we similarly examine acceptance of Blacks: (1) excludes Blacks, (2) includes Blacks, and (3) no stated preference.

Analytic Strategy

We begin by comparing the sample characteristics by both self-identified race and, for Latinos, observed race. Next, we present descriptive statistics of racial homophily preferences, followed by acceptance of Blacks and Whites as dates for the full sample, by race and observed race. To examine whether racial differences are significant once we control for sample characteristics, we next present results from multivariate multinomal logistic regression analyses, focusing first on the Latino sample only and then comparing predicted probabilities of Black and White inclusion among all the racial groups.

Observed Race and Latinos’ Assimilation Patterns

This study’s findings suggest that observed race is a key factor that may lead to divergent assimilation trajectories among the self-identified Latino population. Our unique contribution is to highlight how external racial classification is associated with Latinos’ acceptance (or lack) of others in dating situations. Most existing research suggesting that racial appearance stratifies the assimilation trajectories of Latinos has tended to focus on socioeconomic outcomes, suggesting that discriminatory treatment by others is the key mechanism (Bonilla-Silva Reference Bonilla-Silva2004; Hunter Reference Hunter2005; Telles and Murguia, Reference Telles and Murguia1990). A few other studies have suggested that self-identification choices among Latinos indicate the pursuit of a particular assimilation trajectory (Frank et al., Reference Frank, Akresh and Lu2010; Golash-Boza and Darity, Reference Golash-Boza and Darity2008). In contrast, we show how racial appearance relates to stated acceptance of other racial groups as possible dates, an outcome that potentially impacts mate selection and thus possibilities for intermarriage and marital assimilation. While the Latino daters in this study may be responding to discrimination or lack of acceptance by others with their dating choices, ultimately they are deciding to limit or expand their dating options to particular racial groups; this has implications for assimilation. For example, we find that Latinos who are perceived as White choose to exclude Blacks and include Whites as possible dates at high rates, a decision that may lead to marriage with Whites and to their descendants’ adoption of a White racial identity and categorization. Importantly, current choices can lead to future assimilation regardless of whether that is the intended outcome. We argue that Latino online daters are exhibiting agency in this process by making particular racial dating choices and not others, but also that their agency is shaped by the existing racial structure as evidenced by differences in choices based on how they are racially classified by others.

Thus, this study moves beyond existing research and theories suggesting racial appearance influences assimilation through the impact of differential treatment by others on socioeconomic outcomes (Portes and Zhou, Reference Portes and Zhou1993), by showing that Latinos’ racial appearance also influences their assimilation trajectories through their own decisions to exclude or include members of other racial-ethnic groups as possible dates. While we are not suggesting that the majority of relationships begin online, prior research has shown remarkable consistency between patterns of stated preferences for dates and patterns of interracial pairings (Feliciano et al., Reference Feliciano, Robnett and Komaie2009; Hitsch et al., Reference Hitsch, Hortaçsu and Ariely2010a). This suggests that stated preferences among online daters are capturing preferences that exist among the general population and feed into eventual dating and marriage outcomes. Thus, we argue that these dating choices matter because they capture social distance between groups and influence assimilation—the breaking down of boundaries between groups—in its broadest sense by shaping with whom one marries and/or has children. Through the children born of mixed unions, boundary crossing may occur as descendants of today’s self-identified Latinos may come to identify and be classified as exclusively White or Black (Alba Reference Alba2005). Theoretically, boundary shifting might occur if crossings occurred on a large scale, so that, for example, those who today self-identify and are classified as Latino might be seen and identify in the future as White (Alba Reference Alba2005). Another theoretical possibility is the restructuring of boundaries such that new racial-ethnic groups are formed; for example, self-identified Latinos who are perceived as Black could merge with Blacks into a new racial category that identifies and is seen as “Non-White.”

Our findings suggest that racial appearance is an important factor shaping such possible assimilation outcomes. Among self-identified Latinos who are perceived as White, our findings provide evidence of boundary crossing and eventual assimilation into Whiteness (Yancey Reference Yancey2003). These Latinos accept Whites and exclude Blacks as possible dates to a greater extent than those who are perceived as belonging to other racial groups. Among Latinos who are classified by others as Black, we see evidence of eventual assimilation into Blackness or perhaps developing into a new pan-minority group with Blacks (Carter Reference Carter2005). These Latinos are distinct from those perceived as White or Latino and more similar to self-identified Blacks in that they are much more likely to include Blacks and exclude Whites as dates. Self-identified Latinos whom outsiders observe as Latino seem to fall into a “racial middle” (O’Brien 2008) as a racial group distinct from Whites or Blacks. While these Latinos tend to privilege Whites over other racial groups as acceptable dates, they fall in between “White” Latinos and “Black” Latinos in their levels of excluding Blacks as dates and are far more inclusive of Blacks than self-identified Whites or Asians are. Those Latinos viewed as Latino are also slightly less likely than Latinos perceived as White or Black to exclude other Latinos as possible dates. Thus, it is likely that Latinos who both self-identify and are perceived as Latino are more likely than those who self-identify as Latino but are viewed as White or Black to date (and eventually mate with) other Latinos. When one also considers that most self-identified Latinos in our sample are also perceived by others as belonging to a Latino racial category, separate from Blacks or Whites (72%), this suggests that the “racial middle” may persist for some time. Thus, our findings support the view that assimilation processes vary because Latinos experience racialization differently (Golash-Boza Reference Golash-Boza2006; Golash-Boza and Darity, Reference Golash-Boza and Darity2008). Some who identify as Latino are racialized as White, others are racialized as Black, while still others are racialized as a separate Latino group in the “racial middle” (Bonilla-Silva Reference Bonilla-Silva2004; O’Brien 2008). This view contrasts with perspectives that argue that Latinos as a whole are predominantly assimilating into Whiteness (Yancey Reference Yancey2003) or joining with Asians as a new “non-Black” group (Gans Reference Gans and Lamont1999; Lee and Bean, Reference Lee and Bean2004). We argue that claims about the assimilation trajectories of self-identified Latinos as a whole are problematic given the divergences by racial phenotype.

Why Would Perceived Race Influence Dating Choices?

While our data do not allow us to test the mechanisms through which perceived race shapes dating choices, our findings are consistent with a number of existing theories. According to exchange theory (Davis Reference Davis1941; Fu Reference Fu2001; Merton Reference Merton1941), self-identified Latinos who appear White are more likely to include Whites than other Latinos are because only they have the privilege of being able to exchange their physical capital for a White partner. In contrast, those who are classified as Latino or Black may recognize that their lower racial status constrains them such that they must be more accepting of lower status racial groups in order to date. Self-identified Latinos’ dating choices, according to this perspective, are driven by an acceptance of dominant racial hierarchies in the United States, in which Whites are afforded the highest status.

Another explanation for the link between observed race and racial preferences is through acceptance by co-ethnics: boundaries with other Latinos may be particularly salient for Latinos who do not “appear” Latino and may be assumed to be less authentically Latino by their co-ethnics (Hunter Reference Hunter2005, Reference Hunter2007; Jiménez Reference Jiménez2010); these Latinos may feel more comfortable dating the racial group they appear more similar to. The link may also be through the strength of ethnic identification, which may be weaker among lighter-skinned Latinos, as research suggests it is for light-skinned African Americans (Brown et al., Reference Brown, Ward, Lightbourn, Jackson and Jones1998; Wilkins et al., Reference Wilkins, Kaiser and Rieck2010). Some Latinos may self-identify as Latino, but not feel strong attachments to this identity, and therefore may be more open to dating non-Latinos.

Discrimination or lack of acceptance from other racial groups may also be a mechanism shaping racial choices in dating. Self-identified Latinos who are classified by outsiders as Black or Latino may be more likely than those who are seen as White to have experienced discrimination from Whites. Thus, these Latinos may be more likely to exclude Whites as possible dates as a reaction to negative experiences or because they believe they are unlikely to be accepted by Whites. Such experiences may be particularly pronounced for Latinos who are perceived as Black, which might explain why their acceptance of Whites as dates is relatively low and mirrors that of self-identified Blacks. Thus, the differences we find by racial appearance suggest that Latinos’ agency in choosing dating partners is influenced by the existing racial structure.

The Racial Classification Problem

The findings of this study have implications for the measurement of Latino populations in the United States. Most surveys, as well as the Census, rely only on self-identifications of race, but our research suggests that outsider classification captures a dimension of race that is distinct from self-identification and is independently associated with different outcomes. The Census Bureaus’ current method of separately assessing self-identified race and self-identified Hispanic origin, despite claims that it measures differences by “social” race (Patterson Reference Patterson2001), may inadequately capture the different dimensions of race relevant to understanding the lived experiences of the Hispanic origin population. Our findings illustrate that, not only do outsider classifications of race often not correspond to self-identifications, but in addition, the two dimensions of race yield different pictures of Latinos’ acceptance of other racial-ethnic groups as dates. For example, our sample of self-identified Latinos overall were far more likely to accept Whites as possible dates than Blacks but were also much more accepting of Blacks as dates than self-identified Asians or Whites were. However, these patterns mask some important differences by observed race among self-identified Latinos. We found significant variation by observed race in terms of which racial groups (Whites or Blacks) Latinos were more likely to include as dates, and their degree of acceptance of Blacks and Whites. Thus, while it was the case that those classified by others as Latino or White were more likely to accept Whites than Blacks as dates, the reverse was the case among those classified as Black. Indeed, in terms of acceptance of Whites as dates, “Black” Latino daters were more similar to self-identified Blacks than to “White” Latinos or those perceived as Latino. The greater acceptance of Blacks by self-identified Latinos than by Whites or Asians was also qualified since that acceptance was much greater among “Black” Latinos, followed by “Latino” Latinos, with “White” Latinos only slightly more likely to include Blacks than self-identified Whites did. Thus, we argue that both external assessments of race and self-identity matter in shaping dating choices and subsequent dating and marriage outcomes.

Relying on self-identification data as currently collected by the Census would mask such patterns. Not only has prior research shown that many self-identified Latinos do not identify with a particular racial category based on their physical appearance (Roth Reference Roth2010), the outside observers in this study, consistent with the findings of previous research (Hitlin et al., 2007), viewed Latino or Hispanic as a valid racial category. Although our sample cannot be generalized to the entire U.S. population, our finding that coders perceived only 14% of self-identified Latinos to be White, while 53% of self-identified Latinos on the Census claim a White racial identity, suggests that these two measures often do not correspond (Humes et al., Reference Humes, Jones and Ramirez2011). Indeed, previous research has shown that racial self-identity and outsiders’ assessments of race often conflict (Itzigsohn Reference Itzigsohn2009; Rodriguez Reference Rodriguez2000; Roth Reference Roth2010). Thus, the Census’ current proposal to include Latino or Hispanic as another “race or origin” category would not solve this issue because it does not provide a measure of how others perceive one’s race.

We argue that neither interviewer classifications nor self-identifications of race alone can adequately assess interracial relations through survey research, particularly for Latinos. For example, the patterns of racial inclusion revealed through this study suggest that intermarriage studies that rely on self-identification data alone likely underestimate the degree of intermarriage between self-identified Latinos who appear White and self-identified Whites, and overestimate the degree of intermarriage between self-identified Latinos who appear non-White and self-identified Whites. However, relying only on interviewer classification of race would also be misleading. Our results show, for instance, that the racial preferences of self-identified Latinos who are classified by others as Black differ in important ways from those who self-identify as Black. The former are far less likely than self-identified Blacks to include Blacks as possible dates. Along the same lines, self-identified Latinos who are classified as White are more likely than self-identified Whites to accept Blacks as dates. In order to more accurately capture the nuances of Latinos’ lived experiences of race, including both interviewer assessments of race and self-identifications that include Latino as a category would be ideal. In the absence of interviewer classifications, a question that asks how others view one’s race might be an appropriate proxy.

Future Research

This study suggests a number of directions for future research. To get a better picture of overall assimilation trajectories among the self-identified Latino population, a more representative sample is needed. Here, outsiders classified 72% of self-identified Latino daters as Latino, suggesting the majority of this population experiences racialization as a middle group. However, because of the select nature of the online dating sample, we cannot generalize to how outsiders would racially classify the general population of self-identified Latinos. Even though these internet daters had the option to choose “multiracial” or “other” as their race if they desired, some of the findings might be driven by multiracial Latinos. Although our findings approximate survey situations where respondents are given the option to self-identify as Latino and must identify with only one category, data are needed that can assess whether patterns would differ if daters had the option to identify with more than one racial category.

Our findings show that both self-identity and outsider classification of race clearly influence dating choices, but research that examines the racial appearance of self-identified Blacks and Whites is needed in order to ascertain which dimension of race is more important in shaping dating preferences. Importantly, the mechanisms that we suggest link racial appearance to dating choices are not limited to processes unique only among Latinos. Individuals who self-identify as members of other racial or ethnic groups, such as Black or Middle Eastern, also vary in observed race. Thus, we might expect future research to reveal similar divergences in dating choices by perceived race among these populations. Future research is also needed that would delineate the social and psychological forces behind the patterns we find.

Overall, the patterns here clearly show that perceived race is an important source of heterogeneity within the self-identified Latino population that is associated with differential acceptance of other racial groups as dates. Thus, considering multiple dimensions of race better captures the complexity of how race is experienced by Latinos and how this relates to choices that may lead to divergent assimilation trajectories.