Racial preferences in dating apps: an experimental approach

ABSTRACT

Dating apps can amplify the number and diversity of our potential dates while also enabling us to filter and choose people who are more similar to us. Homogamous preferences and rooted problems like sexual racism can lead users to maintain or even reify a racialized sexual hierarchy that privileges Whiteness. Whether dating apps help build a more racially diverse world or reinforce discrimination is an open question. This study contributes to this debate by analyzing the racial preferences of thousands of users of Tinder, one of the most popular dating apps worldwide. Through a field experiment that displays (otherwise equivalent) female and male Black and White profiles, we compared the choices of people looking for heterosexual and same-gender relationships. The results show that respondents in all groups favor White profiles. Even though previous research sustains that same-gender couples are less racially homogamous, this was not confirmed in this experiment. In fact, the most surprising finding is that the preference for White partners is relatively stronger for lesbian women. However, when predicting the likelihood of a profile being liked, gender and sexual orientation are far more decisive than race. These findings support the need for an intersectional approach toward race, sexual orientation, and gender and to keep studying the effects of dating app use on mating outcomes.

KEYWORDS:

• Online dating
• sexual racism
• interracial relationships
• racial prejudice
• sexual orientation
• field experiment

1. Introduction

Online dating is on the rise and is inserted in a landscape of broad sociocultural, economic, demographic, and technological changes. Our use of dating apps is altering how we find, match, and communicate with our partners (Finkel et al., 2012). These shifts impact people’s most intimate dimensions while potentially modifying social structures by changing couple selection. Rosenfeld et al. (2019) discovered that 39% of couples in the US meet online. These researchers show that the Internet is currently the most popular intermediary, displacing traditional environments to meet a partner like friend groups, family, schools, workplaces, and the neighborhood. Moreover, the COVID-19 pandemic has boosted dating app use (Fortune, 2021). In response to this health emergency, dating apps have played a fundamental role in reconfiguring intimacies, and new online dating practices have proliferated (Gibson, 2021). Thus, there is a need to explore further the social dynamics that take place in these virtual spaces and their consequences.

One of the main differences between offline and online dating is that the latter enables people to access a large pool of potential partners. Generally, in our day-to-day in-person lives, we interact with individuals with similar social characteristics, so we have fewer chances to connect with others who are different from us and build less homogamous relationships. Accordingly, coupling with a person from the same racial background is more usual. Online dating could be changing this pattern since it increases the possibility of meeting strangers without the intervention of previously known social ties, and this seems to have extended the presence of interracial couples (Ortega & Hergovich, 2017). Thomas (2020) found that couples who meet through a dating platform are over 1.5 times more likely to be in an interracial relationship.

However, racially homogeneous patterns also occur in the online dating context, and there is a hierarchy that prioritizes Whiteness. Potârcă and Mills (2015) analyzed online dating data across nine European countries and found that online dating users prefer dating same-race partners while also following racial hierarchies. They also revealed that daters from racially heterogeneous countries like Switzerland tend to date people from minority groups more than in more homogeneous countries like Spain. Christian Rudder (2014), OkCupid’s creator, found that White users are more likely to be messaged and to obtain replies than their racialized counterparts on this platform. Dating platforms use can also reinforce sexual racism and a myriad of discriminatory practices. Since dating apps like Tinder have an interface that emphasizes quick and binary choices based on physical appearances (David & Cambre, 2016), traits like race are especially salient. Other apps directly enable users to filter potential dates according to their race or are especially focused on connecting people of specific racial backgrounds.

To contribute to the debate of whether dating apps facilitate users to increase partner diversity or enable them to perpetuate discriminatory practices, the present study will analyze the racial preferences of Tinder users through a field experiment that compares online daters’ choices with Black and White profiles.

1.1. Racial homogamy and discrimination

According to Kalmijn (1998), partner choices depend mainly on the interplay of three social forces: personal preferences, social group influence, and the opportunities groups have to interact with members from another group (e.g. their availability, group size, and geographical distribution). Dating apps have at least expanded our chances to interact with people from other groups by widening our access to a larger dating market. Therefore, these tools can increase our chances of meeting people of other racial backgrounds.

Since unions like marriages are particularly intimate, homogamy can facilitate estimating how accepted racial minorities are and indicate the social distance among groups within a society (Bogardus, 1968). Interracial relationships are a motor for social change since they can transcend group boundaries through outcomes like marriage and children (Uskul et al., 2007) and decrease racial prejudice (Paterson et al., 2015). Homogamy can have impactful consequences even in the long term; it shapes the characteristics of families and the reproduction of populations at an aggregate level. This can perpetuate inequality within and between generations, potentially impacting the long-run population change (Schwartz, 2013). On the other hand, more exogamous couples can break social boundaries and increase social openness, social mobility, and equality in areas such as education, income, and wealth (Brynin et al., 2008).

Herman and Campbell (2012) observed that White heterosexual individuals are more willing to date people from other racial backgrounds than to intermarry or bear multiracial children. According to this study, positive global attitudes toward interracial relationships do not lead to the personal willingness to cross these racial boundaries. They also found that White women are more likely than their male counterparts to accept interracial relationships for other people but do not want to engage in these relationships themselves. Stillwell and Lowery (2021) propose that gender roles and double gendered standards explain this attitude since heterosexual women face more social penalties for violating norms for intimate intergroup contact, and their role involves preserving the higher status of Whites (e.g. through childbearing).

Dating preferences do not only depend on individual choices but can echo racial inequality and sexual racism. As Callander et al. (2015, p. 1991) define it, ‘sexual racism is a specific form of racial prejudice enacted in the context of sex or romance,’ and it is highly normalized and legitimated as an individual choice on dating apps (e.g. Grindr). Sexual racism is manifested in a continuum that ranges from subtle to explicit. This ‘personal preference’ discourse is part of the neoliberal narrative that resonates with dating apps’ design and use, legitimizing extended cultural assumptions that influence people’s desires (Robinson, 2015). These preferences reflect how socially constructed preferences mirror and perpetuate an unequal social system (Bourdieu, 2001). Following the feminist claim, the personal is political. Our intimate lives are deeply intertwined with other macro levels where racism is prevalent.

1.2. Contextualizing ‘race’ and its relevance for interracial relationships and online dating

Race is a socio-historically constructed categorization framework, and this concept is increasingly under debate. Researchers in Europe often prefer using ‘ethnicity,’ which includes factors like culture. Wimmer (2008, p. 973) defines ‘ethnicity’ as a ‘subjectively felt sense of belonging based on the belief in shared culture and common ancestry.’ In this paper, ‘race’ refers to a classification system that groups people with similar phenotypic characteristics (which is also socially constructed and not straightforward). How we interpret phenotypes impacts how we categorize each other and has implications regarding the inequality people might experience depending on the category where they are placed (Feliciano, 2016). We want to focus on race since traits like skin color and their social importance are especially relevant in this study due to the role of photographs, looks, and attractiveness on dating apps. We use words like ‘racialized’ to highlight that these classifications based on ‘race’ are part of socio-historical and cultural processes instead of neutral categorizations based on inherent characteristics (Ahmed, 2002).

In today’s highly image-focused dating apps, race is an especially salient trait. Ranzini (2019) suggests that since dating apps offer an environment with reduced cues to evaluate potential partners, race becomes explicitly present as a factor. In a lab experiment with fictitious profiles, Ranzini and Rosenbaum (2020) found that all the participants rated Caucasian-looking profiles as more attractive and were more willing to date them, and attractiveness strongly predicts partner selection inonline dating. Racial stereotypes might be playing a role in this context. Visual representations – like the profile pictures displayed on dating apps – of racialized groups might switch on stereotypes automatically, even when the viewers self-report egalitarian attitudes (Payne et al., 2005).

On the other hand, Whiteness is invisibilized and de-racialized, assumed to be ‘just human,’ the ordinary, the standard, and the norm (Dyer, 1997). The notions associated with race are being challenged. They are dynamic, context-specific, change over time, and are not homogeneous. However, they impact our preferences and can harm people, particularly those further from the top of the hierarchies of desire. Beliefs like racial stereotypes and racialized beauty standards impact interracial attraction and provoke harmful consequences on racialized minorities’ sexual and romantic encounters – including discrimination and fetishization – leading these daters to experience lower self-worth and self-esteem (Silvestrini, 2020). Likewise, racialized groups are also stratified and (dis)advantaged based on how light their skin is due to colorism (Hunter, 2007).

1.3. Sexual orientation and racial preferences

Although research on racial homogamy is mainly based on heterosexual relationships, another essential piece of this analysis is sexual orientation. Same-gender couples are less racially homogamous than heterosexual ones (Lundquist & Lin, 2015; Rosenfeld & Kim, 2005; Schwartz & Graf, 2009). Various complementary mechanisms are proposed to explain this pattern. People looking for potential same-gender partners have access to fewer candidates, and this scarcity might lead them to form more heterogeneous couples. Another reason might be that queer people tend to live in larger cities and racially diverse neighborhoods (Chauncey, 2008). Horowitz and Gomez (2018) propose an identitarian explanation. They found that when people identify themselves actively with a member of the gay, lesbian, and bisexual community, they have more chances to engage in interracial couples. As the underlying mechanism for this phenomenon, these researchers propose an identitarian override based on a shared status as outsiders.

There are also gendered differences among people looking for same-gender relationships. Research shows that gays are more likely to interracially cohabitate and marry than lesbians (Schwartz & Graf, 2009). Based on an analysis of an anonymous online dating site, Lundquist and Lin (2015) suggest that while White men’s higher number of interracial relationships occurs due to more constrained dating markets, their lesbian counterparts have more open racial preferences. Rafalow et al. (2017) found that lesbians conform to less racialized and gendered normative ideals on Match.com. One explanation they propose to explain this draws on the work of authors like Rich (2007): considering that since lesbians might be more invisibilized, they are freer than gay men to be transgressive. Nonetheless, Rafalow et al. (2017) also found that both lesbian and gay groups still were racially homogamous, especially in the case of White users. At the same time, racialized people were far more likely to exclude their own group than their White counterparts, potentially reflecting internalized racial oppression. When stating their racial preferences in their profiles, online daters often offer gendered explanations based on ideal masculinities and femininities for their choices. In line with these results, empirical research on heterosexual online dating has found that Asian males and Black females are the most highly excluded groups (Robnett & Feliciano, 2011; Rudder, 2014). One reason might be that their social representations deviate respectively from hegemonic masculinity and femininity. Therefore, gender, sexual orientation, and race need to be analyzed together. Following Collins and Bilge (2020, p. 2), an intersectional approach is needed since the power relations based on these variables ‘are not discrete and mutually exclusive entities, but rather build on each other and work together; and that, while often invisible, these intersecting power relations affect all aspects of the social world’.

Researchers who study dating apps have also found that racialized users often experience racist behavior on dating apps. It is common to see openly sexually discriminatory comments that are presented as personal choices (e.g. ‘Not attracted to Asians’) on profiles on apps for men looking for men like Grindr (Callander et al., 2012, 2015). Although there is significantly less research about queer women’s experiences with dating apps, researchers have found that they also experience racial discrimination in this environment (Li & Chen, 2021). Simultaneously, as Han and Choi (2018) affirm in their paper Very Few People Say ‘No Whites’, racialized people are not just sexual racism ‘victims’, but also actively confront these hierarchies of desire.

1.4. The Spanish context

Most of the literature cited in this paper is based in the United States. We need to broaden our understanding of these topics by studying other countries. In the Spanish context, the elites have avoided debates about race and this country’s colonialist past, leading to racial discrimination that those who are racialized based on their physical appearance suffer the most (Flores, 2015). There is scarce research about interracial unions in Spain. However, this information is particularly relevant since, from the nineties, an increasing number of people from diverse geographical areas and cultures immigrated to this country. This increased Spanish society’s diversity and the growth of marriages and families from different nationalities and racial backgrounds (Albert-Guardiola & Masanet-Ripoll, 2008). In 2021, in 14,75% of marriages in Spain, one of the members did not have a Spanish nationality while the other did (INE, 2022b). There is no official data about the percentages of interracial marriages since official institutions like the National Institute of Statistics (INE) do not ask about race. Most studies are based on the nationalities of the members of the couples and other dimensions of homogamy, like education (Cortina-Trilla, 2009, 2016; García-Román, 2021).

On the attitudinal level, the approval towards interracial relationships seems high. According to a Special Eurobarometer about discrimination in the European Union (2023), 86% of Spanish people would feel totally comfortable if one of their children was in a romantic relationship with a Black person, in contrast with the EU average (68%). Only the Netherlands (92%) and Sweden (92%) had higher acceptance percentages than Spain. In contrast, Hungary (37%) and Bulgaria (24%) had the lowest. Spanish respondents showed the highest acceptance toward their children having a romantic relationship with White people (96%). Unluckily, the Eurobarometer does not ask about the participants’ dating preferences or if they have experienced race-based dating discrimination. Regarding attitudes towards sexual minorities, according to the same Eurobarometer, Spain shows more favorable attitudes than the average in the EU. While 80% of the respondents in Spain would feel comfortable if their children were in a love relationship with a person of the same sex as their child, the EU average is 59%. Spain is only behind the Netherlands (91%), Sweden (89%), and Denmark (83%). Romania (23%) and Bulgaria (16%) show the lowest acceptance levels. However, it should be considered that factors like social desirability might bias these answers, and this is not a measure of peoples’ real attitudes and behaviors. There are differences between these acceptance ratings and the experiences that racialized people report.

On the other hand, the Council for the Elimination of Racial or Ethnic Discrimination surveyed racialized people who live in Spain (Consejo para la Eliminación de la Discriminación Racial o Étnica, 2020). They found that they perceive the most significant discrimination in these fields: access to the housing market (31%), establishments or public spaces (30%), and the workplace (26%). The groups who perceived the most discrimination based on their skin color and physical appearance are people from Sub-Saharan Africa (82%) and Roma people (71%). Unfortunately, this study did not ask about discrimination in their romantic experiences. These discriminatory practices occur in a context where racist hate speech grows in Europe and is spread and amplified online with tools like fake news (Gamir-Ríos et al., 2021). The rise of radical right-wing political parties across this continent, including Spain (Mendes & Dennison, 2021), might also have an impact on interracial dating since their discourses are heavily based on components like national identity and anti-immigration attitudes. Therefore, we can also expect discrimination in intimate relationships since all these areas are interconnected. Considering the impact dating and coupling have on our social structure and quality of life, widening our knowledge about these dynamics in unexplored world areas is essential.

1.5. The present study

In line with prior research and literature on homogamy, since the Spanish population is predominantly White and considering that racial discrimination exists in this society, we expect that most people will prefer racially homogamous partners choosing Caucasian profiles more often. In accordance with the literature on racial homogamy and sexual racism, Hypothesis 1 expects that dating app users prefer White over Black profiles when looking for both heterosexual and same-gender relationships.

We also expect to find differences based on the sexual orientation of the respondents. Hypothesis 2a expects users looking for same-gender relationships to be more likely to like and initiate conversations with Black profiles on Tinder compared to users looking for heterosexual relationships. Existing literature sustains that same-gender couples are less racially homogamous. The causes behind this fact are that they look for partners in a thinner market, they usually live in more racially heterogeneous urban areas, and they might hold more egalitarian values and are more willing to cross social boundaries (Horowitz & Gomez, 2018; Rosenfeld & Kim, 2005; Schwartz & Graf, 2009). Since these patterns have been demonstrated for same-gender relationships, this hypothesis predicts that respondents looking for same-gender relationships will be more willing to express interest in the Black experimental profiles compared to their heterosexual counterparts. Previous research also shows that gays conform more to racialized, gendered ideals than lesbians and are more likely to have race-based dating preferences (Rafalow et al., 2017). Therefore, we propose hypothesis 2b: the Black female profile is expected to receive more likes and messages than the Black man in the same-gender condition.

We want to emphasize that the term ‘same-gender’ will be used to mention the ‘women seeking women’ or ‘men seeking men’ options in this app. It should be noted that users looking for these options might identify themselves with other terms like bisexual or even heterosexual. It is also possible that they do not identify themselves as women or men. However, this category is defined as ‘same-gender’ following the binary design this app had when we conducted this experiment.

2. Method

Identifying, measuring, and registering discrimination in dating is challenging, especially in its most covert expressions. Assuming that racial equality is a widely shared value, it can be expected that people do not want to admit discriminatory thoughts or practices to researchers or to themselves due to social desirability reasons. Then, techniques based on self-reporting are not suitable for this study, even when the identity of the respondents is anonymous. This research will gather empirical evidence through field experiments that facilitate comparing subjects who are equal in all their characteristics except for one (e.g. race). This approach combines experimental methods with field research, simulating real-world interactions while retaining key experimental features that help draw causal inferences (Pager, 2007). Field experiments have effectively measured racial discrimination in social sciences. Moreover, this method has uncovered discriminatory practices applied to contexts like the workplace (DiStasio & Larsen, 2020), the housing market (Ahmed & Hammarstedt, 2008), and public spaces (Zhang et al., 2022).

Now, we have the technological tools to go beyond traditional field experiments. The digital age amplifies our opportunities to collect and analyze social data efficiently, gives us access to massive samples, and enhances our creative possibilities to explore real-life contexts (Salganik, 2019). For example, Edelman et al. (2017) showed that applications from guests with typical African American names are more likely to be rejected on Airbnb. Doleac and Stein (2013) found on Craigslist that people are more likely to respond to an ad selling an iPod if a White hand holds it than to an identical ad with a Black hand. Applying these designs in a digital platform enables us to observe discrimination directly, unlocking insights that other methods would not have uncovered. Although digital field experiments are complicated to design and implement rigorously to guarantee all the variables are equal except the analyzed one, they help us simply and quickly conduct experiments with sizable samples anywhere, at any time.

2.1. Profile creation

This experiment followed a 2 (gender: female, male) x 2 (race: Black, White) x 2 (sexual orientation: gay/lesbian, heterosexual) design. Four fictitious and analogous profiles were created on Tinder. These profiles included: a Black woman, a White woman, a Black man, and a White man. They were designed to keep all the variables equal except one (race, gender, and sexual orientation in each case). The profiles’ attractiveness level, picture layout, age, and bios were comparable. To control that their attractiveness was equivalent, the pictures were taken from the Face Research Lab London Set (DeBruine & Jones, 2017). This dataset contains images of subjects who self-reported their age, gender, and race. All the photographs were taken with equal clothes, lighting conditions, and layouts. Figure 1 shows pictures of the four profiles. 2,513 people evaluated the attractiveness of all the subjects. Participants rated their attractiveness from ‘much less attractive than average’ to ‘much more attractive than average’ on a scale from 1 to 7. For this experiment, the attractiveness rating of the chosen subjects varied from 3.55 to 3.86. This ensured that participants in this experiment perceived these fictitious profiles as equally beautiful. The chosen photographs portrayed the subjects staring at the camera in a frontal position and looking to their right quarter. All the photographed subjects received the instruction to smile. However, there are slight differences in their facial expressions. The images were edited with Photoshop to make them more similar to pictures that users upload to these dating apps (e.g. the default light gray standard background was replaced with a white room with plants). Since people portray their hobbies on these online sites, a third picture was extracted from Unsplash, a stock photo website. This last image portrayed the back of one subject with the same skin color in a natural environment (e.g. hiking in a forest). Afterward, four accounts were created on Tinder. The names of the profiles were taken from the list of the most frequent names among the Spanish population born in the decade 1990, elaborated by the Spanish Institute of Statistics (INE, 2022a). All of them were 26 years old and had equivalent bios that mentioned that they were from Madrid and displayed common hobbies like the types of films and food they enjoyed. Premium subscriptions were purchased to access the app’s special features, like seeing the total number of likes every profile received and maximizing their visibility.

Figure 1. Pictures of the four profiles.

This experiment took place in Madrid because it is the capital of Spain, and its large population is more diverse than the population of other areas of this country. 14,02% of people living in Madrid have a foreign nationality (INE, 2022c). Although it is not free from discrimination, Madrid is one of the friendliest and safest destinations for LGBTIQ+ people in the world. This capital hosts the biggest Pride celebration in Europe and one of the most internationally known queer – primarily gay – neighborhoods worldwide, Chueca. Therefore, more people might be using dating apps and feel free to engage with the profiles on the same-gender condition in Madrid. This also means we could possibly detect more discriminatory practices in other parts of Spain, like regions with a more homogeneous population and less open-minded attitudes, like rural areas.

2.2. Experimental design

The chosen dating site was Tinder, a geolocation-based app. All the profiles were located in different equivalent spots in the center of Madrid. In this app, users can browse pictures of different profiles. They can like them or reject them. If two users like each other, it is a match. Then, they can start a conversation. A filter with an age range of 20–36 years was set to browse the respondents. Our distance preferences were up to 159 km, so we interacted with profiles of dating app users within that distance radius. These were all the preferences we set up. All the data was collected between April 2022 and May 2022. Each pair was compared by gender with the same sexual orientation for a week (e.g. the two female profiles in the same-gender setting). Every created profile browsed 650 profiles and liked 80% of them. This means that they swiped right (liked) on 520 profiles. Therefore, 4,160 profiles were liked in total, counting the four profiles in all the groups. A simple Python script was followed while using the app to simulate selecting 80% of the profiles randomly. To prevent bias, we used a script that randomly selected the Tinder users’ profiles that our experimental profiles should like instead of relying on the researchers’ criteria. This random selection also helped avoid setting off any alarms the app might interpret as suspicious activity. For instance, if we had simply liked the first 520 profiles we came across, the app may have flagged our actions as non-human activity and banned our accounts.

Profiles that did not show pictures of their faces or were offering commercial services were excluded. Firstly, the created profiles sought potential heterosexual relationships (this dating app’s design is quite binary). After the week and the data were gathered, the profiles would look for potential same-gender partners. Later, this process would be replicated with the profiles from the other gender. These profiles did not talk to those who liked them or sent them messages. This design enabled thousands of people to view these profiles.

2.3. Data analysis

The number of likes, matches, and messages the profiles received were registered daily in a database. The identity of the respondents was anonymized. Subsequently, all this quantitative data were analyzed using SPSS.

2.4. Ethical considerations

The Ethics Committee of the authors’ university approved this experiment.¹ The ethical dilemmas the chosen method entails are: 1) the design impedes the respondents from giving their informed consent to participate, and 2) the profiles created did not belong to real users, which involves deceiving the respondents. Nonetheless, this was the only way to obtain unbiased and real replies that would not be attainable through techniques like interviews and artificial environments like a lab. Since the contact respondents had with these profiles was minimal, and the volume of interactions users have on these platforms is usually high, this research’s risk is minimal. Additionally, the respondents’ privacy was respected. All their data is anonymous, making it impossible to identify them. Therefore, the welfare and rights of the respondents were respected at any moment throughout the research process. Applying the principle of beneficence established by the Belmont Report, the social benefits of shedding light on racism in this emergent dating context and the benefits of this research justify these small potential risks.

3. Results

Thousands of people saw these profiles and preferring White profiles over Black ones became a clear and extended pattern. Tinder does not show the total number of views every profile had, so calculating the total number of respondents is impossible. This estimation is based on the number of likes the profiles got. The most selected profile, the White heterosexual woman, received 7,471 likes. However, men outnumber women as dating app users (Fiore & Donath, 2005; Fiore et al., 2010), and we can expect far fewer respondents in the same-gender conditions. The outcomes obtained by each profile in terms of matches, likes, and conversations are shown in Table 1. It was not possible to calculate the likes percentage because we do not know the number of Tinder users who saw every profile.

Table 1. Outcomes for each profile: matches, conversations, and likes.

Profile	Right swipes	Matches	Conversations	Likes	Match percentage	Conversations percentage
White heterosexual woman	520	282	170	7,471	54.23%	32.69%
Black heterosexual woman	520	229	94	4,925	44.04%	18.08%
White gay man	520	149	59	210	28.65%	11.35%
Black gay man	520	109	44	168	20.96%	8.46%
White lesbian woman	520	94	33	132	18.08%	6.35%
Black lesbian woman	520	39	12	52	7.5%	2.31%
White heterosexual man	520	39	6	41	7.5%	1.15%
Black heterosexual man	520	27	4	29	5.19%	0.77%

The first hypothesis predicted that dating app users prefer White to Black profiles in heterosexual and same-gender conditions. To test this hypothesis, two linear probability models were fitted. When the outcomes of a treatment are binary (e.g. receiving a like or not), this technique is highly advantageous due to its flexibility and simplicity. Linear probability models are the simplest way to estimate correlations easily and facilitate interpreting the results directly in terms of probabilities (Gomila, 2021). This technique offers better results in terms of interpretability compared to similar techniques, such as logistic models. In this analysis, both linear probability models had a single independent variable that indicated whether the subject was Black. Each instance of liking a profile was considered a separate data point, and the binary dependent variable was taken to be the outcome of this. In Model 1, the fact of a match arising was considered the dependent variable (coded by 1 for a match and 0 for no match); in Model 2, the fact of a conversation being initiated by the liked profile was considered the dependent variable (coded by 1 for an initiated conversation and 0 for no conversation). These different indicators point out the possibility of being chosen by the respondents. The models were fitted in SPSS using least squares with the complete dataset collected in the experiment. The significance of the coefficients was evaluated using a T-test. Table 2 shows the results of these models.

Table 2. Results of the linear probability models for the likelihood of receiving a match or message as a Black (vs. White) profile.

Model	Coefficient	P-Value
Model 1: Matches	−.077	p < .001
Model 2: Messages	−.055	p < .001

The models predict that, on average, a Black subject will get 7.7% points fewer matches and 5.5% points fewer conversations. Considering that, on average, these profiles got 23% matches out of all likes, this is quite a strong effect. These results confirm hypothesis 1.

Hypothesis 2a predicted that users looking for same-gender relationships are more likely to like and initiate conversations with Black people on a dating app compared to their heterosexual counterparts. Hypothesis 2b expected that the Black female profile would receive more likes and messages than the Black male profile in the same-gender condition. To support these hypotheses, four additional linear probability models were fitted for each separate group of sexual orientation and gender. This was done similarly to Model 1 in the previous paragraph (with matches being the dependent variable) but fitted four times on different subsets of the data. The results were as follows (see Table 3):

Table 3. Results of linear probability model 1: probability of getting a match for a Black (vs. White) profile.

Profile	Coefficient	P-Value
Lesbian female	−.106	p < .001
Heterosexual female	−.102	p < .001
Gay male	−.077	p = .004
Heterosexual male	−.023	p = .127

The second set of hypotheses is not supported. In the gay male group, men still had liked and texted the White profile significantly more than the Black profile. In the female lesbian condition, respondents showed a significant preference for the White female profile. In terms of the size of the regression coefficient, these preferences are even stronger in the gay and lesbian condition than in the heterosexual one. Remarkably, the only condition in which the effect is not statistically significant is the heterosexual male. This is partly due to the heterosexual males getting fewer matches, which causes the sample size to be smaller.

Similarly to Model 2, four more models were fitted with conversations being the dependent variable. This process was applied to each separate group. These were the results (see Table 4):

Table 4. Results of linear probability model 2: probability of getting a message for a Black (vs. White) profile.

Profile	Coefficient	P-Value
Lesbian female	−.040	p < .001
Heterosexual female	−.146	p < .001
Gay male	−.029	p = .120
Heterosexual male	−.004	p = 0.526

The effect regarding conversations in terms of the size of the regression coefficient is larger in the female heterosexual condition than in the lesbian. However, it is weaker relative to the total percentage of conversations of the corresponding White profile (for the White heterosexual profile, 32.69% of the right swipes lead to a conversation, while for the White lesbian profile, this percentage is only 6.35%, as shown in Table 1). The effect in both the gay and heterosexual male groups is not significant. Especially in the heterosexual condition, very few conversations occur (only 6 for the White profile and 4 for the Black profile), which makes it impossible to draw any definitive conclusions.

In conclusion, the experiment results do not indicate less discrimination in the female lesbian condition compared to the heterosexual. In the male group, the difference is not significant (partly due to a lack of samples). Hence, our second set of hypotheses must be rejected.

Finally, to enhance the interpretability of the results, we translate the coefficients into odd ratios to show how White profiles are preferred. White profiles had more likes in every situation. In the heterosexual group, the White man had x 1.41 likes, and the woman had x 1.52 likes. In the same-gender condition, the White man got x 1.25 likes while the White woman x 2.54 likes. Regarding the matches, in the heterosexual group, the White man obtained x 1.44 matches, and the White woman 1.23. In the same-gender condition, the White man had x 1.37 matches, and the White woman 2.41.

Regarding the number of messages, the White-person profiles also received more texts in all situations. In the heterosexual group, the White man got x 1.5 messages more and the White woman x 1.81 messages. In the same-gender condition, x 1.34 messages were sent to the White man while x 2.75 messages were written to the White woman. Finally, we want to remark that, in the heterosexual condition, all four women who started a conversation with the Black male profile and one out of the six who messaged the White male reported in their bios that they were trans women.

4. Discussion

Dating apps significantly impact people’s lives, and the outcomes of these interactions can have a lasting effect on our social structure. Previous research shows that couples who meet online are more likely to cross racial boundaries and be more diverse in other aspects (Ortega & Hergovich, 2017; Thomas, 2020). In this study, White profiles were significantly more likely to be liked and to receive messages compared to Black profiles, confirming hypothesis 1. These results might be linked to the racial composition of the Spanish population and coincide with other studies that show how daters prefer people from their same racial backgrounds and that expose a racial hierarchy that privileges Whiteness (Potârcă & Mills, 2015; Rafalow et al., 2017; Robnett & Feliciano, 2011). Although dating apps increase the chances of meeting people from different backgrounds that we generally do not meet in our in-person social circles, racialized preferences are perpetuated in these digital spaces. Far from using digital scenarios to erase barriers, people can also pursue their desires in these virtual contexts, experiencing fewer inhibitions without filters and feeling freer from the pressure of external opinions. The obtained results in this study are aligned with the previous literature about racial homogamy, racial preferences, and sexual racism.

Hypothesis 2a expected that users looking for same-gender relationships would be more likely to like and initiate conversations with the Black profiles, and it was rejected. This result contradicts previous research about homogamy in same-gender couples (Horowitz & Gomez, 2018; Lundquist & Lin, 2015; Rosenfeld & Kim, 2005; Schwartz & Graf, 2009). One common explanation for the propensity of same-gender couples to be interracial is that they have access to fewer potential partners, so they tend to cross racial boundaries more. Since dating apps increase their options, scarcity could stop being a reason to engage in interracial relationships for people looking for same-gender partners. Instead, heterosexual men are the ones who encounter scarcity because there are more male than female users on dating apps, and women tend to be more selective in this context. As other researchers have found (Timmermans & Courtois, 2018), female profiles received far more likes and messages in the heterosexual condition than male ones. This ultra-competitive environment for men looking for heterosexual relationships might make them less racially selective in this initial step of the dating process. However, it should be noted that Duguay (2019) found that queer women also report facing scarcity using mainstream dating apps.

Hypothesis 2b expected that the Black female profile would receive more likes and messages than the Black man in the same-gender condition. Although previous researchers like Rafalow et al. (2017) found that lesbians conform less to racialized, gendered ideals than gay men and are less likely to have race-based dating preferences, the most surprising finding of this study is that the preference for White subjects is relatively stronger for lesbian women compared to all other groups. Further research is needed to explain this result. This discrimination is especially harmful considering the triple jeopardy and multiple minority stress Black lesbian women experience due to this intersection of variables (Bowleg et al., 2003).

People usually interpret racial affective and sexual preferences as personal preferences, not perceiving them as part of a racist attitude (Callander et al., 2015). Nonetheless, these seemingly individual choices are socially constructed and embedded in a broader racially-based hierarchy prioritizing Whiteness. These attraction preferences are interconnected with the gendered discriminations that Black people face in other spheres like the workplace, the housing market, and public spaces. This situation can even have lasting consequences on their well-being and health, hurting their self-esteem and self-worth and leading them to experience a wide range of negative emotions. These adverse outcomes are multiplied when they are also LGBTIQ+ (Balsam et al., 2011; Calabrese et al., 2015; English et al., 2018) and/or experience another kind of discrimination.

Additionally, although singleness is on the rise, couples are privileged in this heteronormative couple-centric society (Budgeon, 2008). Although getting many likes in an app does not necessarily translate into having more chances to have a successful stable relationship, obstacles to getting a partner can be related to other damaging outcomes like economic hardships (e.g. difficulties buying a house with a single income). Racialized sexual stratification has structural roots and a systemic impact. Social structures and culture shape sexual and affective desire and mirror social inequalities, while attraction patterns reflect and maintain them.

Apart from the results of this experiment related to race, gender disparities were far more remarkable and outweighed the racially-based differences. For example, in the heterosexual groups, while the Black woman received 4,925 likes, the White man was liked 41 times. From the evolutionary perspective (Buss, 1989; Trivers, 2017), women tend to be more careful with their choices because they invest much energy in pregnancies and parenting. Instead, men are motivated to maximize their chances of passing their genes to the next generations. From a sociological and feminist point of view, in Western hetero-patriarchal society, although (most) women are also able to have multiple relationships and are supposed to enjoy them, they are judged harshly for engaging freely in these activities due to the existing double standards (Armstrong et al., 2010). Even though this enormous asymmetry in likes can be interpreted as women having more power in the dating field, these differences can be due to other factors like traditional dating scripts. The initial advantage that one app user can have in this early dating stage can be reversed in later courtship steps.

Another remarkable finding is that all the female participants who initiated a conversation in the heterosexual condition with the Black male profile and one out of the six who talked to his White counterpart were trans women, according to their bios. One plausible explanation is that since trans women are commonly attacked and discriminated against, suffering a tremendous amount of violence in all dimensions of their lives (Faye, 2021), dating apps can offer them a safe space where they can disclose their identity upfront, and filter out people who would not accept them (Albury et al., 2021). Researchers have found that trans women use dating apps to avoid harassment and build intimate connections but also experience transphobia and feel fetishized in this environment (Albury et al., 2021). Considering the discrimination they face, they might need to be more active than cisgender women to interact with other users on a dating app.

Dating apps’ dynamics need to be explored from an intersectional approach, examining the complex interactions among variables like gender, sexual orientation, and race. These racial orders and gendered dynamics are location and time-specific, take place in a determined context, and are fluid. Although the results of this study do not point in this direction, there can be room for social change that could be technologically enhanced. Users are also active in choosing how they build their relationships and use technologies, altering this complex picture where (not) racialized bodies, attraction patterns, social hierarchies of desire, gendered dynamics, and technological developments intersect.

4.1. Limitations and future research

Field experiments enable us to examine respondents’ behavior unobtrusively in a real-world scenario and help us prevent biased replies due to causes like social desirability. However, this study’s experimental design does not facilitate knowing other relevant variables like the respondents’ nationality and personality traits. Unfortunately, calculating these matches’ racial homogamy rate was impossible since we did not register the race of the respondents. Without knowing the respondents’ race, we can not determine if the results mostly indicate a racial homogamous preference of White people preferring people of their race or if racial minorities also share this preference. Future studies could examine racial dating preferences within specific racial groups in Spain and add more racial backgrounds to similar experiments.

In this experiment, the attractiveness of all the profiles was equivalent. However, slight differences between them, such as facial expressions, could influence our findings. We could also expect other results if there were differences in the profiles created, like changes in their age and bios. Another interesting potential experiment would be creating profiles with different attractiveness levels to examine if respondents would select Black profiles more often if they were perceived as more beautiful than White ones. Moreover, the partner choice process depends on multiple factors. Apart from race and attractiveness, there are also fundamental variables like socioeconomic status, education, personality traits, and lifestyle. Following the social exchange theory (Blau, 1964), people evaluate their (potential) partners’ desirability through a subjective calculus based on perceived rewards and costs. In this case, it could be explored if a characteristic perceived as an advantage, like a high education or attractiveness level, compensates for a race that respondents find less attractive. The data collection lasted one week for each experimental condition. We acknowledged that we could have expected more likes and messages if we had waited longer (e.g. one month). Considering the large sample that participated in this experiment, we do not expect these differences to impact the study’s results. Regarding our statistical analysis, performing interaction analyses would have provided a more adequate test of our hypotheses. However, the sizable variation in the number of matches and conversations across different conditions (as shown in Table 1) disrupts the interaction analysis results, leading to erroneous interpretations.

Although this design facilitated exploring users’ revealed preferences, this method does not permit exploring the reasoning behind their choices or the effects these dating practices have on them. Complementing this study with other qualitative techniques that ask the respondents directly would have enabled delivering a more detailed, thicker description and a more complete picture. We could have explored other issues like the fetishization of racial minorities, which was not feasible with this experimental design as we were not aware of the respondents’ intentions. With our present design, we could not ask respondents to identify their gender and sexual orientation. In most cases, we only knew the respondents’ condition (e.g. man looking for heterosexual relationships). That does not necessarily reflect the respondents’ self-identification. For instance, a non-binary bisexual person could appear in the women seeking women category (which we labeled same-gender).

It would have also been insightful to analyze and compare each profile’s conversations if they had replied to the respondents’ messages to obtain a nuanced qualitative analysis. Nevertheless, from an ethical perspective, this prolonged interaction with these research subjects would constitute an unacceptable risk for the respondents. Another limitation is that since the algorithms that partially direct users’ interactions in this app are unknown, it is possible that they could have altered this experiment in unexpected ways. These results could vary if, instead of exploring Tinder, we would have used other apps with other affordances and used by other user types. Future collaborations with dating apps are necessary to keep broadening our knowledge about this topic.

Finally, online dating platforms are used for the first steps of the coupling process. Every match is not necessarily followed by a date. Some users might engage in interracial casual relationships but not want to continue with them in the long term. Previous literature suggests interracial couples decrease through from dating to cohabitation to marital unions (Blackwell & Lichter, 2004). In this study, following up on what happens later is not possible. These findings should be complemented with longitudinal data about the outcomes of the matches that take place through dating apps. It would be helpful to keep deepening our understanding of racial preferences in queer female relationships due to these surprising results and the lack of research to explain them. Future researchers could draw international comparisons and explore the intersectionality between race and other variables like age, attractiveness level, body size, gender identity, disability status, and socioeconomic background.

5. Conclusion

Although dating apps increase our chances of meeting people from different racial backgrounds, these results reveal that respondents looking for same-gender and heterosexual relationships preferred White profiles. This study has obtained empirical data from a field experiment to uncover and compare racial preferences. This method enabled direct exploration free from biases like social desirability. Although previous research states that same-gender couples tend to be less racially homogamous and have less pronounced racial preferences, this was not confirmed in this experiment. Surprisingly, the preference for White partners is relatively more prominent in the lesbian female condition. However, gender and sexual orientation are more important than race in predicting the likelihood of an online dating user being chosen. Additionally, since most research about topics such as interracial relationships and sexual racism is based in the US, this contribution helps build a bigger picture by adding evidence about the Spanish context, which is underexplored. This study also advances the understanding of queer women in dating apps, which is crucial because research on online dating is mostly heterosexual or gay male-centered, and research on racism in the queer community is more focused on gay men. Future research should explore the potential of dating apps for social change and building a less segregated and unequal society.

Acknowledgments

The authors express their gratitude to Joana Maria Pujadas Mora, Paul Puschmann, and the anonymous reviewers for their insightful and constructive feedback.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data will be made available on request.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Notes

1. This study was approved by the ethics committee of the institution where the first author was affiliated during the development of this study. She is currently affiliated with another institution.