How Do Intersections of Sexual Orientation and Race Affect Impressions of Women in a Counter-stereotypical Job Context? A Comprehensive Test Manipulating Individuating Competence Information

The Stereotype Content Model (Fiske et al., 2002) provides a theoretical framework of how stereotypes of social groups can be described using the two central dimensions of warmth (also referred to as communion) and competence (more broadly referred to as agency, see for example, Abele et al., 2016). Agency comprises traits related to task-competence, along with those related to assertiveness and dominance. Extending the Stereotype Content Model, the newer ABC model (A. Koch et al., 2016) proposes that beliefs (regarding progressiveness vs. conservativeness) are a third central evaluation dimension next to agency and communion.

Most stereotypes are ambivalent (Cuddy et al., 2011), including gender stereotypes (for a review see Ellemers, 2018): Traditional women are associated with high warmth and low competence due to women’s ascribed social roles (Eagly & Steffen, 1984; Eagly & Wood, 2012). However, leadership positions generally have masculine connotations, so that they are perceived to be inconsistent with prescriptive and descriptive stereotypes of traditional women (Heilman, 2001). According to Heilman’s Lack-of-fit model (also see Eagly & Karau, 2002), this makes women appear less suited for leadership positions (Badura et al., 2018; Castano et al., 2019), less likeable, and less desirable as supervisors in a male-typed job context compared to men (Heilman & Okimoto, 2007). Thus, women (implicitly assumed to be White and heterosexual) applying for a male-typed job should be judged relatively low in competence and consequently hireability.

Such gender stereotypes are not applied to all women alike (e. g., reviewed by Steffens & Viladot, 2015). Women differ based on their origin, their educational background, and their age, to name just a few dimensions. This variation is hinted to by research on subgroups within the Stereotype Content Model (Brambilla et al., 2011; Clausell & Fiske, 2005; Klysing et al., 2021; Kotzur et al., 2019). The stereotype content regarding specific subgroups is often different from respective overarching categories (Fiske et al., 2002). Group stereotypes are most commonly applied to prototypical applicants (i. e., conforming to the norm of being White and heterosexual, Klysing, 2023; Purdie-Vaughns & Eibach, 2008). For example, when compared to traditional women, associations are reversed for business women, who are perceived to be competent but cold (Asbrock, 2010; Eagly & Steffen, 1984; Eckes, 2002; Fiske et al., 2002).

The term intersectionality, coined by Black feminist scholars and activists (Combahee River Collective, 1978), highlights analytical approaches that consider the co-construction, meaning, and consequences of (belonging to) multiple oppressed social groups (Crenshaw, 1989). Such an approach emphasizes that categories such as gender, ethnicity, or sexual orientation simultaneously influence how an individual is perceived (Cole, 2009). Belonging to multiple marginalized groups does not necessarily lead to disadvantages (Pedulla, 2014; Purdie-Vaughns & Eibach, 2008; Strinić et al., 2020). Instead, findings are mixed (for a review, see Ball et al., 2022). We therefore believe it is fruitful to use the Stereotype Content and Lack-of-fit models to derive specific predictions of intersectional discrimination. So what specific stereotypes are associated with different groups of women?

Implicit Inversion Theory (Kite & Deaux, 1987) is a lay theory according to which people often assume that lesbian women are similar to straight men who are often ascribed high competence, but low warmth (Fiske et al., 2002; for results on partial gender inversion, see Klysing et al., 2021). Corroborating such subgroup stereotypes, some experiments found that lesbian women were more likely to be judged as competent but cold (Asbrock, 2010; Mize & Manago, 2018) or as more competent than heterosexual women (Niedlich et al., 2014, 2022; Niedlich & Steffens, 2017). However, different lesbian subgroups were also judged differently in terms of competence and warmth (Brambilla et al., 2011; Geiger et al., 2006), which might explain contradictory findings. In one US-American study, lesbian women were also perceived to be progressive (Jensen et al., 2022).

In contrast to such findings corroborating the Stereotype Content Model, where lesbian applicants were perceived as more competent than heterosexual applicants, meta-analytic data from Western Europe and North America showed that lesbian and gay applicants generally encountered hiring discrimination; but discrimination was lower in high-skilled (vs. low-skilled) occupations, and lesbian women were less discriminated against than gay men in gender-neutral and male-typed jobs (Flage, 2019). Lesbians were perceived as more masculine (Niedlich et al., 2014), thus resembling straight men, and were therefore seen as a better fit for male-typed jobs or a worse fit for female-typed jobs (Drydakis, 2015). One hypothesis one can deduce based on these findings is that lesbian women applying for a male-typed job are judged higher in competence and hireability than heterosexual women.

Even though stereotypes of Black people in general were found to be ambivalent with medium competence and warmth (Eckes, 2002; Fiske et al., 2002), and Black professionals were stereotyped as more competent and less warm than Black people in general (Fiske et al., 2002), there are also differences in terms of gender: more unique stereotypes regarding Black women than men were found (Ghavami & Peplau, 2012). Gendered stereotypes have nuanced impacts on Black women in the working context depending on whom they are compared with. In the U.S., Black women and White men were perceived as more dominant (which is one facet of agency) than White women and Black men and thus perceived as more suited for leadership positions (Livingston et al., 2012). However, Black women in leadership positions were rated similarly to Black men and White women on career success, but continued to be rated more negatively than White men (Rosette & Livingston, 2012). Since being White is deemed an essential attribute of the prototypical business leader (Rosette et al., 2008), this finding is not surprising. However, Black female managers’ abilities were not downplayed as was the case for Black male managers (Greenhaus & Parasuraman, 1993). A field experiment in Europe showed that Black women applying for service and administrative jobs were discriminated against compared to White and Asian women both for male- and female-typed positions, but when applying for male-typed jobs Black women were somewhat less discriminated compared to White women (Di Stasio & Larsen, 2020). Despite stereotypes of Black women depicted as less intelligent and skilled (Settles, 2006), and their contributions being overlooked in non-work settings (Sesko & Biernat, 2010, 2018), they were also stereotyped as assertive and dominant, which are aspects of agency important for leadership recognition (Livingston et al., 2012; Rosette et al., 2016). Thus, whereas no clear-cut pattern of discrimination of Black women was found, on the basis of the latter findings, one can deduce the hypothesis that Black women applying for a male-typed job are judged higher in competence and hireability than White women.

The difference between Black and White women could be even more prononunced for Black lesbians. Studies focusing on the gendered aspects of racialization (Ghavami & Peplau, 2012; Johnson et al., 2012; Schug et al., 2015) have shown that Black individuals were perceived as more masculine than other racialized groups (Hall et al., 2015). Consequently, one can deduce the hypothesis that a Black lesbian woman is considered as a relatively suitable candidate for a male-typed job. However, this could depend on the culture in which the study is conducted.

Comparing stereotypes of Black people in Germany and the U.S., Black people’s assumed foreignness and refugee status emerged as the most striking difference (Ball et al., 2022). In Germany, Black gay men specifically are seen as gay and no longer as prototypical of their racialized group. It is unclear whether Black lesbian women (associated with a refugee-status in Germany, see Pilot Study 1, Online Supplement, OS) would be seen as lacking competence and warmth or whether their sexual orientation would shift perceivers’ impressions towards higher competence. There is evidence that foreigners – along with Turkish people and Muslims (Asbrock, 2010) – and refugees more generally (Kotzur et al., 2019) are perceived as rather cold and incompetent, whereas both lesbian women and gay men are seen as relatively high in competence (Asbrock, 2010). Yet, how Black people in Germany are stereotyped has hardly been examined (for an exception, see Koopmans et al., 2019).

It could be that discrimination depends on the origin attributed to a Black person within Europe, which was found for male applicants in some studies (Derous et al., 2016; Weichselbaumer & Schuster, 2021). Thus, it is important to disentangle discrimination based on ancestry (foreign-sounding name) and on phenotype (race) in research on racial discrimination in Europe (Polavieja et al., 2023). For example, names perceived as German as compared to African could provoke more positive impressions. Moreover, in a U.S.-based study, disclosing a non-normative sexual orientation buffered negative consequences for Black gay men (Wilson et al., 2017). It stands to question whether these results extend to female applicants. In Germany, not much research on Anti-Black racism has been conducted (Aikins et al., 2021; Ball et al., 2022). One recent mixed-methods survey focusing on experiences of discrimination by Black and Afro-descendant people in Germany found that 85% of participants (N = 4098) had been discriminated in the job-context and that 29% (N = 2586) indicated working in a job below their qualifications (Aikins et al., 2021). In other words, survey responses of Black people’s perceived discrimination in Germany appear to contradict the hypotheses that we derived from theoretical models of impression formation (albeit these were specific to male-typed jobs); and not all studies testing the Stereotype Content and Lack-of-fit models yielded consistent findings. An additional factor that could shed light on these complicated patterns of findings is individuating information.

In job application processes, as a rule, additional information about applicants is present. Such individuating information has the potential to reduce discrimination based on group membership (Agerström et al., 2012); to generally increase callback rates for all applicants (Veit et al., 2022); or to even lead to a reversed pattern of discrimination (Rosette & Tost, 2010). However, the role of individuating information in mitigating hiring discrimination is unclear. Since group stereotypes encompass different levels of competence and warmth (Fiske, 2012), they could interact with the respective individuating information in systematic ways (Niedlich et al., 2014). For instance, a lesbian or Black applicant may need to demonstrate only a small amount of additional information describing her as assertive (in line with stereotypes) to be perceived as non-traditional (for a woman) and thus highly agentic. In contrast, a White heterosexual applicant may need to present more of that additional (counter-stereotypic) information to be perceived as non-traditional. On the contrary, it could be that these assertive lesbian or Black applicants are perceived as less warm than heterosexual White applicants (for whom being warm is taken for granted because it is stereotypical). For example, one recent study showed that demonstrations of female-typed behavior affected decisions (i. e., exclusion, promotion) regarding Turkish heterosexual women more negatively than German heterosexual women, whereas the contrary was found for demonstrations of male-typed behavior (Ball et al., 2024). To test the hypothesis that individuating information interacts with group stereotypes in affecting impressions, one could vary individuating information indicating different levels of competence to determine under which conditions impressions are influenced by social categories.

Taken together, stereotypes at intersections of gender, race, and sexual orientation have complex effects in different contexts, and previous findings on women’s job-related discrimination have not yielded a consistent pattern for who is favored for a highly-qualified male-typed job. Moreover, particularly regarding race, there is a scarcity of quantitative research adopting an intersectional approach (Else-Quest & Hyde, 2016). Possible interactions with individuating information have also been neglected, and too few different stimuli have typically been used.

Against this backdrop, the aim of the present experiment was to make the following original contributions. First, we examined job-related impressions of women at the intersection of race (Black vs. White) and sexual orientation (lesbian vs. heterosexual). We used a male-typed job-context (i. e., surgery) that should be most counter-stereotypical for prototypical women (i. e., heterosexual, White). Second, we investigated how impressions depend on individuating competence/warmth information, testing whether prototypical female applicants need to present more information on their competence than non-prototypical applicants to prove themselves quite different from the stereotype of traditional women. In pursuing these aims, we went beyond typical experiments by not presenting only one specific application profile and selectively manipulating the social category, but using a whole set of different pictures of applicants (Prager et al., 2018; Wells & Windschitl, 1999).

To that end, we asked participants to review 32 fictitious application profiles for a male-typed leadership position with photos of lesbian or heterosexual Black and White women containing self-descriptions of four traits. To manipulate individuating information on competence/warmth, traits varied from rather warm (three warm and one competent trait) to competent (four competent traits). The specific traits used were also varied to allow generalizability. After seeing each applicant, participants were asked to indicate inferred competence, warmth, conservativeness, and decide on the applicant’s exclusion versus retention in the application pool. We pre-registered all of the following hypotheses prior to the start of data collection (https://aspredicted.org/JZK_PQB). We report all measures, manipulations, and exclusions in the present manuscript. The experiment was part of a series that had received approval by the local ethics committee.

Regarding Black or lesbian applicants: Hypothesis 1a assumed that lesbian applicants are rated more competent than heterosexual applicants (Niedlich et al., 2014, 2022; Niedlich & Steffens, 2017) and are thus more likely to be retained in the application pool (H1b). Because Black women were perceived as more assertive and not being disadvantaged in leadership positions (Livingston et al., 2012; also see Rosette & Livingston, 2012), we expected Black female applicants to be rated as more competent than White female applicants (H2a) and more likely to be retained in the application pool (H2b).

Regarding intersections of race and sexual orientation: We expected that non-prototypical applicants, i. e. Black and heterosexual, Black and lesbian, White and lesbian, would be rated more competent than prototypical applicants, i. e. White and heterosexual (H3a) (Strinić et al., 2020). Likewise, non-prototypical applicants would be more likely to be retained in the application pool than prototypical applicants (H3b). White lesbian applicants could be seen as more assertive than White heterosexual applicants due to their sexual orientation, whilst Black lesbians could be seen as even more assertive based on their sexual orientation and race. Yet, because they are stereotyped as being less intelligent than White women (Ghavami & Peplau, 2012; see Pilot Study 1 in OS for details), we did not assume substantial differences between their evaluation and that of White lesbian applicants. In other words, we expected that all non-prototypical applicants are seen as a good fit for the vacant position.

Considering the role of individuating information on competence/warmth (e. g., Fiske & Neuberg, 1990; Niedlich et al., 2022), we finally assumed that (H4a) non-prototypical applicants would be considered more competent even at lower levels of competence than prototypical applicants and that (H4b) non-prototypical applicants would be more likely to be retained in the application pool than prototypical applicants, even at a lower level of competence. In other words, White heterosexual women would need to demonstrate more competence (counter-stereotypical for them) than women belonging to any other group. Because of the preeminent role of inferred competence in male-typed job settings (Badura et al., 2018; Castano et al., 2019; Koenig et al., 2011), we did not formulate directed hypotheses on warmth ratings, but explored them. Similarly, we explored how conservative (A. Koch et al., 2016) applicants were rated.

Pilot studies. We conducted five pilot studies to obtain first evidence of the ideas the present experiment is based on (for details, see OS). Pilot Studies 1 (N = 24) and 2 (N = 43) assessed culturally known stereotypes of straight/gay and male/female Black people in Germany and compared them to stereotypes sampled in the U. S. A main finding was that Black individuals were stereotyped more frequently according to the stereotypes generated in Germany than stereotypes identified in the U.S. Of note was that both straight and gay/lesbian Black people were perceived as foreigners.

In Pilot Study 3 (N = 92), we obtained first evidence for the hypothesized interactions between social category and individuating information for inferred warmth and competence of female applicants in a male-typed job. We found that when provided with one particular profile of individuating trait information, German heterosexual women appeared significantly less competent than lesbian women (similar to Niedlich & Steffens, 2015). This was not observed with other profiles, indicating that individuating information can affect social-category based discrimination. Pilot Study 4 (N = 14) confirmed that the manipulation worked we use in the main experiment: competence as expressed by a target applicant’s traits was related to differing inferred competence. Pilot Study 5 (N = 14) established that six different pieces of additional, work-related information communicated applicants’ sexual orientation in a subtle way (e. g., “Employee X is always available, even on weekends, because her wife usually works on weekends.”).

Participants. The main online survey ran from 24.08.2022 until 30.08.2022 with participants being recruited by an external German company (“Clickworker”, https://www.clickworker.de/clickworker/) and directed to an online-platform (www.soscisurvey.de). Their monetary compensation was oriented at the level of the minimum wage and given to participants who fully completed the experiment. Predefined inclusion criteria were a minimum age of 18 and being employed at least part-time. In addition, we used gender and age quotas to obtain a sample representative of the German population regarding these two criteria. Our final sample consisted of 212 participants with M_age = 41.43 years (SD = 13.32). 108 were female (50.8%) and 104 were male (49.1%). Among 352 participants who answered the attention check (“Please choose the rightmost answer.”), 125 participants were excluded due to an incorrect response as pre-registered; 13 additional participants were excluded because of repeated participation. On average, participants took 16.05 minutes (SD = 4.60).

Sexual orientation was assessed on a five-point Likert-type scale (from 1 “lesbian/gay” to 5 “heterosexual”); 89.2% identified as heterosexual (M = 4.82, SD = 0.64). In terms of formal education, the majority had a university degree (N = 90) or completed an apprenticeship (N = 44) and 56% indicated having at least some prior experience with personnel selection (M = 2.18, SD = 1.02, on a four-point scale from “1 – not at all”, 32%; 2 – 30%; 3 – 26%; to “4 – a lot”, 12%). For historical reasons, racial self-identification is not assessed in Germany in surveys, but most likely, our sample was almost entirely White.

Design. We used a 2 (sexual orientation: lesbian/heterosexual) x 2 (racialized groups: Black vs. White) x 4 (number of competence-associated traits indicating: “competent”, “rather competent”, “neutral”, “rather warm”) within-subject design. The cover story presented female applicants for a male-typed job as leading surgeon posting their documents at an internal application platform. The presentation order of the targets was randomized to exclude order effects; assignment of the profiles regarding sexual orientation to the photos was fixed because race (obviously) also was.

To account for possible participants’ idiosyncratic variance connected to faces, names, and sexual orientation, and variance connected to stimuli, we analyzed the data via a hierarchical multilevel model. To determine sample size we used a Shiny App (Westfall et al., 2014). For a counter-balanced design with 24 photos of Black and White target applicants, 205 participants provide sufficient power (1- β = .80) expecting an effect of d = 0.4 and α = .05. Sensitivity analysis showed that with the final N = 212 and 24 stimuli we could detect medium-sized effects of d = .51 with a power of 1- β = .95 and α = .05.

Materials. Each participant saw 32 photos of smiling White (n = 20) and Black (n = 12) women from the Chicago Face Database (Ma et al., 2015). Among these, eight photos of White heterosexual women served as fillers increasing the number of the numerical majority (for details, see OS). We matched the photos regarding attractiveness and masculinity within six clusters, so that that each cluster contained a heterosexual White and Black applicant and a lesbian White and Black applicant. Sexual orientation was manipulated via notecards by a former supervisor on applicants’ profiles giving additional information concerning their private life, mentioning their female or male partner (e. g., “husband [wife] works from home so moving would not be a problem”). We chose German-sounding names (Wiktionary, 2023) being associated with comparable age, attractiveness, and intelligence (Kleen & Glock, 2020; Nett et al., 2020; Rudolph et al., 2007) to rule out effects of names (Koopmans et al., 2019; Weichselbaumer & Schuster, 2021).

In detail, each participant saw 32 photos, of which 24 photos depicted an equal number of White and Black targets with fixed name, race, and sexual orientation that were counterbalanced regarding individuating competence information. In each cluster of four photos, every possible combination of race and sexual orientation occurred once. Each photo was paired once with each of the four competence levels across the four questionnaire versions. In the “competent” condition, four adjectives were randomly drawn from the pool of competence-related terms. In the “rather competent” condition, three competent adjectives plus one warmth adjective were selected. In the “neutral” condition, two adjectives each were drawn from each pool (e. g., confident, ambitious, helpful, empathic). Finally, in the “rather warm” condition, three warmth adjectives and one competent adjective were drawn. Filler profiles were assigned a fixed competence level (two profiles per level) across all questionnaire versions. To make the application profiles look as realistic as possible, checked boxes indicated that the applicants had successfully uploaded their CV, a photo, a certificate of their final exam in medicine, and a specialist certificate for training as a surgeon. On purpose, photos were rather small to allow easy detection of racialized groups without drawing attention to individual facial features, and information diagnostic of high competence was omitted (e. g., obtained grades).

Procedure. After informed consent, participants were randomly assigned to one questionnaire version. Subsequently, we asked for socio-demographics regarding gender, age, and occupation to ensure the correct quota. All eligible participants were then presented the cover story and asked to assess 32 applications in their imagined role as a recruiter at a large, international hospital group, within which internal applicants were applying for a team-leader position. The exclusive presentation of female applicants was justified by the company’s efforts to promote equality and thus the explicit search for a female team leader. We reminded respondents to exclude some applicants because they could not interview everyone.

Then, participants saw the web design of the alleged internal application portal from the applicants’ point of view to make sure they had seen all 16 adjectives for the manipulation of individuating competence information. Afterwards, they saw each of 32 application profiles. In a self-paced manner, they clicked to rate, on the subsequent page, each applicant on competence, warmth, conservativeness, and hireability. In the beginning, participants evaluated two filler profiles to get used to the procedure, including an attention check. All filler profiles had a fixed position in the sequence. The target profiles were shown in randomized order within and partly also between the clusters. To assess competence (α = .98) and warmth (α = .98), we used one item each of the facets as developed by Abele et al. (2016), namely competence (“competent”), assertiveness (“confident”), warmth (“friendly”), and morality (“reliable”) with a 1-7 Likert-type scale (e. g., “very little competent – very competent”). Conservativeness was rated on one item, “1 – modern” to “6 – traditional” (A. Koch et al., 2016). Finally, we assessed applicants’ hireability by asking participants to choose whether they wanted to keep the applicant in the application pool or exclude her. We collected socio-demographic information on sexual orientation, education, German language skills, and experience in personnel selection before debriefing participants.

Multilevel analyses. Given the non-hierarchical clustered data structure (Barker et al., 2020) with stimuli (i. e., photos) that are nested at once in social identity categories (i. e., racialized group and sexual orientation) and in competence levels, we created cross-classified multilevel models (CCMM) (Goldstein, 1994), so that the combination of the photos with competence levels (that were counterbalanced) was modeled at Level 1, while the photos with fixed race and sexual orientation were modeled at Level 2, as were the competence levels. All analyses were performed using RStudio (Team, 2022), with the lme4 package for CCMM (Bates et al., 2015). For details, see OS.

Manipulation checks. Before fitting CCMM, we checked whether targets with varying individuating competence information were indeed perceived as differing in competence. To that end, we fitted a MLM with competence condition as a Level 1 categorical predictor variable and photos at Level 2 (for details, see OS). The “rather warm” (i. e., gender-stereotypical) category was used as the reference group for all comparisons. As expected, and shown in Figure 1, applicants who described themselves as “rather competent” were inferred more competence than applicants in the reference group “rather warm”, γ₁₁ = 0.12, p = .007. Unexpectedly, applicants in the condition “neutral” were rated lower in competence than applicants in the reference group, γ₁₁ = -0.20, p < .001. Regarding warmth, we found that applicants in the condition “rather competent” (γ₁₁ = 0.23, p < .001) and “competent” (γ₁₁ = 0.20, p < .001) were inferred more warmth than applicants in the reference group. In short, applicants’ selecting more competence-related traits led to their being perceived higher not only in competence, but also in warmth.

The correlation coefficients between inferred warmth and competence yield an explanation for the counter-intuitive finding that selecting fewer warmth-related traits went along with more inferred warmth: Competence and warmth ratings were highly correlated (r = .49, p < .001; for correlations between warmth and competence ratings for applicant subgroups, see OS, Tables OS 10-13). Apparently, participants formed generally a more or less positive impression of a given applicant and then rated them accordingly on all presented traits. Supporting this idea, correlations with inclusion/exclusion decisions (point-biserial correlation with Holm-adjustment) showed that the warmer or more competent an applicant was rated, the more likely she was kept in the pool of applicants, with two large effects of p = .54 and p = .56 respectively, p_s < .001: Warmth information played a similar role as competence information, despite the job context in which competence should be most relevant.

Additionally, we examined participants’ reaction times when looking at the application profiles (for details, see OS). Supporting the idea that the most stereotypical (fastest) impressions were formed of White heterosexual targets, participants spent least time looking at White heterosexual applicants’ profiles, χ²(3) = 24.49 (p < .001, Friedman test). In detail, participants spent on average less time looking at White heterosexual (M = 14.05, SD = 10.26) than at Black heterosexual (M = 17.52, SD = 30.45), White lesbian (M = 15.73, SD = 14.47), or Black lesbian applicants (M = 16.29, SD = 18.41; see Discussion for interpretation).

Next, hypotheses were tested by first comparing applicants depending on sexual orientation, then on racialized group perception, and then on the interaction of both identities. In each case, each of the four competence conditions was taken into account, too, if feasible, and respective interactions. Targets in all four social-identity categories were rated as rather warm and competent (see Figures 2-3), and mostly invited for an interview (Figure 4), with OR = 1.98 [95% CI: 1.71, 2.31]. To anticipate, hardly any of the central hypothesized effects could be statistically confirmed. In short, we found significant main effects of competence levels both for the DV inclusion and for the DV warmth, indicating that generally applicants whose self-description implied more competence (“rather competent”, “competent”) were perceived as warmer and were more likely to be retained than those in the reference group. Additional significant findings occurred for conservativeness ratings: Here, we found an interaction that was similar to the one we had hypothesized regarding hireability. More prototypical applicants had to demonstrate their counter-stereotypicality more strongly such that applicants expressing more competence were rated as less conservative than applicants expressing more warmth. White lesbian applicants also needed to demonstrate more warmth (counter-stereotypical of them) rather than competence to be perceived as less conservative than White lesbian applicants expressing more competence.

Hypotheses tests are summarized in the following sections with Table OS 6 showing intraclass correlations (indicating how similar two values are within the same Level 2 unit – in this case, regarding each photo and competence conditions, see Eid et al., 2015) and random effects of all calculated models, as well as the mean intercepts of all null models, and Table OS 7 showing an overview of the fixed effects of the predictors, inferential statistics, and effect size estimates for all hypotheses tests.

Comparison of lesbian and straight applicants. In Model 1.1, testing Hypothesis 1a (expecting lesbians to be rated as more competent than heterosexual women), the intercept of (both Black and White) lesbian applicants was non-significantly higher than that of heterosexual applicants, pseudo-R² < .0001. Similarly, lesbian applicants were non-significantly estimated to be higher in warmth than heterosexual applicants, pseudo-R² = .0004 (Model 2.1). To test Hypothesis 1b that lesbian applicants were more likely to be retained in the applicant pool than heterosexual applicants, a GLMM was set up which, in addition to the cross-classified data structure according to Model 2, also took into account the dichotomous response format (with 0 = exclusion, 1 = inclusion). Whereas the intercept of lesbian applicants was descriptively higher compared to heterosexual applicants, with OR = 1.06 and the confidence interval of the OR comprising 1, this difference was not significant. Thus, Hypothesis 1b was not supported.

Comparison of Black and White applicants. Model 1.2 on inferred competence indicated that, in contrast to our expectations, Black (heterosexual and lesbian) applicants were descriptively rated less competent than the reference group of White applicants (heterosexual and lesbian), but this difference was not statistically significant, pseudo-R² = .0004. Accordingly, Hypothesis 2a was not confirmed. Whereas Black applicants were descriptively rated as warmer than White applicants, this difference was also non-significant, pseudo-R² = .0004. Similar to competence, the hireability of Black compared to White applicants (Model 3.2) was rated non-significantly lower, OR = 0.99.

Comparison of prototypical and non-prototypical applicants. According to Hypothesis 3a, non-prototypical applicants (Black and heterosexual, Black and lesbian, White and lesbian) should be rated as more competent than prototypical applicants (White and heterosexual). To confirm this hypothesis, significance of both main effects of sexual orientation and race in Model 1.3 was required. White lesbian applicants were descriptively rated a bit higher in competence than White heterosexual applicants, and Black heterosexual and lesbian applicants, a bit lower, but the analyses revealed no significant main effects of the two predictors, pseudo-R² < .0001, and thus no evidence for H3a. While Black heterosexual applicants were rated as warm as White heterosexual applicants, both White and Black lesbian applicants were rated slightly warmer than White heterosexual applicants, but this was not statistically significant, pseudo-R² < .0001. Regarding hireability, White lesbian and Black heterosexual applicants were a bit more likely to be retained in the applicant pool than White heterosexual applicants, and Black lesbian applicants a bit lower, but not significantly so: All confidence intervals for sexual orientation, OR = 1.10, 95% CI [0.79, 1.53], for race, OR = 1.02, 95% CI [0.74, 1.43], and for the interaction of sexual orientation and race, OR = 0.93, 95% CI [0.58, 1.48], included 1 and therefore, Hypothesis 3b was not supported.

Comparison of applicants depending on the competence condition. Hypothesis 4a assumed that non-prototypical applicants would be rated as more competent than prototypical applicants already when providing fewer traits indicating competence (and more traits indicating warmth, Model 1.4). Consistent with our expectation, applicants presented with more traits indicating competence were also rated somewhat more competent (except for the competence condition “neutral”), but this effect was not statistically significant. Consistent with previous analyses, the intercept increased non-significantly for White lesbian compared to heterosexual applicants and decreased non-significantly for Black (lesbian and straight) applicants. As none of the interactions was significant, pseudo-R² < .0001, Hypothesis 4a was not confirmed. Comparing warmth ratings depending on competence condition, sexual orientation, and race, in contrast to our expectations, the more competent applicants’ self-descriptions were, the warmer they were rated, with a small effect, pseudo-R² < .0001 (replicating the Manipulation Check). Rather warm heterosexual White applicants represented again the reference group whose intercept of warmth decreased for the condition “neutral” as the number of traits indicating competence increased, while it significantly increased for the conditions “rather competent” and “competent”. White lesbian and Black straight applicants were descriptively, but non-significantly, estimated to be warmer compared to heterosexual applicants, while Black lesbian applicants were rated slightly less warm compared to the reference group. No significant interaction terms emerged.

Hypothesis 4b assumed that non-prototypical applicants were more likely to be retained in the applicant pool than prototypical applicants even when presenting with fewer traits indicating competence. Since the interaction model testing this assumption did not converge, we changed the cross-classified data structure to a classic hierarchical multilevel model without competence conditions at Level 2, as well as a corresponding null model. The interaction model (Model 3.4) showed that the intercept increased significantly for competence conditions “rather competent” and “competent” compared to “rather warm”, OR 1.48, 95% CI [1.16, 1.90], p = .02 and OR 1.72, 95% CI [1.35, 2.20], p = .001: Generally applicants portrayed as more competent were more likely to be retained in the applicant pool than applicants portrayed as warmer. Thus, Hypothesis 4b was not supported.

Given the number of null findings regarding the hypothesis tests, one might wonder whether participants even noticed the different social groups the applicants belonged to. Applicants’ sexual orientation was not highly salient, but manipulated by notecards referring to a male/female partner. Exploring conservativeness ratings, we fitted CCMM for null and predictor models, and a classical hierarchical MLM for the interaction model with a 2 Level structure because the model with the cross-classified data structure did not converge. Table 1 illustrates all intraclass correlations and random effects, Table 2 presents the intercepts as well as statistical characteristics of the fixed effects for all conservativeness models (see Figure 5). Model 4.1 showed that (Black and White) lesbian applicants were rated as significantly less conservative than heterosexual applicants. When we included only race as predictor (Model 4.2), Black applicants were descriptively, but non-significantly, rated lower on conservativeness than White applicants. Model 4.3 with both predictors included simultaneously showed a significant main effect of sexual orientation, such that White lesbian applicants were perceived as significantly more modern than the reference group of heterosexual White applicants (see Figure 5). Model 4.4 tested possible interaction effects between degree of expressed competence, race, and sexual orientation. Applicants who presented themselves as “rather competent” (i. e., in contrast to traditional gender stereotypes) were perceived to be significantly less conservative than when they presented as “rather warm”. In contrast, “competent” White lesbian applicants were rated as more conservative than White lesbian, “rather warm” applicants (i. e., those who contrasted stereotypes of lesbians). This result approached significance with p = .050, pseudo-R² = .0078. In sum, findings on conservativeness demonstrate that participants’ ratings indeed depended on applicants’ social identities.

Neutral competence condition. To see whether results regarding competence ratings and exclusion decisions differed when reducing part of the complex design, we conducted additional analyses only considering the neutral competence condition. The results mirrored the above (null) results (see Tables OS 8 and OS 9).

Participant features. One could imagine that older or male participants discriminate more against minority participants, whereas younger or female participants prefer a more diverse team (e. g., Everly et al., 2015; Steffens & Wagner, 2004). Therefore, in additional analyses, we included participants’ age and gender separately as predictors in interaction with targets’ sexual orientation and race within each of the models that we had fitted for hypotheses tests. The patterns of results (available upon request from the first author) hold except for conservativeness ratings and exclusion decisions: The finding that “competent” White lesbian applicants were rated as more conservative than “rather warm” White lesbian applicants was statistically significant for average-aged male participants, γ₁₁ = 0.30, SE = 11, t = 2.69, p = .007. Further, male participants excluded both lesbian applicants more often than straight ones, γ = -0.29, SE = 0.08, z = -3.48, p < .001, OR = 0.75 [95% CI: 0.63, 0.88] and Black applicants more often than White ones, γ = -0.22, SE = 0.08, z = -2.57, p = .010, OR = 0.81 [95% CI: 0.68, 0.95].

Classic ANOVAs. In most of the research that we based our hypotheses on, ANOVAs instead of MLAs had been computed. For direct comparison, we also ran 2 (race: Black vs. White) x 2 (sexual orientation: straight vs. lesbian) repeated-measures ANOVAs on the facets of competence (respectively warmth) included as repeated-measures factor (competent, assertive for competence; reliable, friendly for warmth). Regarding overall competence, we found a main effect of race, F(1, 211) = 5.433, p = .021, partial η² = .025, with White applicants being rated as more competent than Black applicants. An interaction effect of race and sexual orientation for the facet competence, F(1, 211) = 5.904, p = .016, partial η² = .027, indicated that White lesbian applicants were rated as more competent than other applicants. For the facet assertiveness, a main effect of race showed that White applicants were rated as more assertive than Black applicants, F(1, 211) = 7.244, p = .008, partial η² = .033.

Overall warmth ratings showed a main effect of sexual orientation such that lesbian applicants were rated as warmer than heterosexual ones, F(1, 211) = 7.635, p = .006, partial η² = .035. Two main effects emerged for the facet friendliness. Lesbian applicants and Black applicants were rated as being friendlier than White or straight applicants, F(1, 211) = 7.291, p = .007, partial η² = .033 for race, and F(1, 211) = 18.590, p < .001, partial η² = .081 for sexual orientation. Regarding the facet reliability, we found that White applicants were rated as being more reliable than Black applicants, F(1, 211) = 4.068, p = .045, partial η² = .019.

Regarding conservativeness ratings, not only a strong main effect of sexual orientation emerged (as in the multilevel analysis), F(1, 211) = 44.959, p < .001, partial η² = .176, but also a main effect of race, F(1, 211) = 9.264, p = .003, partial η² = .042. Both lesbian applicants and Black applicants were perceived to be more modern than straight or White applicants.

Taken together, ANOVA results deviated from the above-described multilevel analyses with several significant main and interaction effects. In other words, accounting for the random variance, as we did, led to much smaller estimates of effect sizes than conventional ANOVAs. In contrast, conducting classic McNemar-tests for the forced-choice question on inclusion in the applicant pool did not yield differing results for comparison of straight White and Black applicants or lesbian White and Black applicants, χ²(1) < 0.01, p > .99, w = 0.00 and χ²(1) = 1.633, p = .201, w = 0.09, respectively.

In line with the Lack-of-fit model (Heilman, 1983), discrimination of women is more pronounced in male-typed than in neutral or female-typed jobs (e. g., A. J. Koch et al., 2016). However, contradictory findings have been obtained regarding identity intersections with sexual orientation and race, showing both hiring discrimination (Flage, 2019; Niedlich et al., 2022; Rosette & Livingston, 2012) and no discrimination (Di Stasio & Larsen, 2020; Livingston et al., 2012; Niedlich & Steffens, 2017; Rosette et al., 2016). Additionally, hardly any research on the discrimination of Black women exists in Germany. Generally, experiments have typically been limited in their scope, often only manipulating targets’ social category while holding constant the qualification profile and job context. Thus, stimulus sampling was typically wanting (for a noteworthy exception, see Veit et al., 2022). Moreover, numerous older experiments favored analyses of variance over multilevel analyses, likely leading to an overestimation of effects, as suggested by our results in direct comparison.

In a large within-subject experiment with an age-representative gender-balanced sample of the German population who worked at least part-time and mostly had some experience with personnel selection, we used photos of 24 different female targets crossed with four different competence-warmth levels. We examined the influence of individuating information regarding competence on job discrimination of applicants differing in sexual orientation and race to test whether applicants’ chances of being invited for an interview depended on their intersectional positionality combined with individuating information. Even though participants noticed the subtle differences between applicants regarding warmth and competence, contrary to our expectations, non-prototypical applicants were not perceived to be substantially more (nor less) competent than prototypical applicants. Non-prototypical applicants were also not substantially more (nor less) likely to be retained in the application pool. Regarding White lesbian applicants we found an interaction with targets’ expressed competence. That is, they had to reveal less additional information (being “rather warm” vs. “rather competent”) than prototypical applicants to prove their progressiveness. Thus, findings show that perceived progressiveness is not only relevant in forming impressions of groups of people (Jensen et al., 2022; A. Koch et al., 2016), but also of individuals. Yet, when it came to competence ratings, non-prototypical applicants did not have to reveal less (nor more) additional individuating information than prototypical applicants to prove their competence contrasting other findings (Agerström et al., 2012; Veit et al., 2022). Against our expectations, inferred warmth ratings showed that targets were perceived as warmer, the more competent their self-description was. Targets were rated high in both warmth and competence with significant positive correlations among competence and warmth, different from ambivalent stereotypes postulated in the Stereotype Content Model (Fiske et al., 2002). Even though previous findings indicate that lesbians have sometimes better chances to be hired (Niedlich et al., 2022), there is also evidence of similar hiring opportunities (Niedlich & Steffens, 2015) or even discrimination against lesbian applicants in low-skilled jobs (Flage, 2019). Accordingly, for both lesbian Black and White women we did not find that they were inferred more (nor less) competence or given better (or worse) chances to be invited compared to heterosexual applicants.

Overall, applicants’ social group membership thus had little influence on competence or warmth impressions and on decisions to retain applicants in the application pool, but some influence on conservativeness ratings. These partial null findings on discrimination based on sexual orientation, race, and their intersection were obtained despite substantial participant and stimulus samples (especially given the within-subject design) and despite the manipulation of individuating information on competence-warmth only via applicants’ self-descriptions (i. e., remaining ambiguous).

Notably, we did detect some significant differences in inferred competence as a function of the experimental competence condition; that is, when applicants described themselves with more traits signaling competence (“rather competent” vs. “rather warm”) they were perceived to be more competent. Likewise, they were more likely to be retained in the application pool. Since we did not find this difference across all competence conditions, we interpret these findings as an indication that the manipulation worked in principle, but that it was rather weak given the clearly male-typed, highly qualified position as leading surgeon and applicants’ appropriate qualification. The influence of stereotypes and ensuing discriminating occurs mostly in ambiguous situations (Dovidio & Gaertner, 2000, 2004; Isaac et al., 2009), which we tried to create by not providing unambiguously high qualifications (Koopmans et al., 2019; Livingston et al., 2012; Weichselbaumer & Schuster, 2021), such as testimonials or a reference letter. Anyway, it seems that the job context exerted a strong effect, emphasized by the generally high competence (and warmth) ratings (i. e., all above 5, scale 1-7). This was the case even though only very little information was given about applicants (just enough to demonstrate that they had the required education – omitting grades, and some prior work experience), indicating the importance of context factors. One might thus find discrimination when applicants clearly present not so good qualifications or in earlier stages of hiring at entrance levels (Thijssen et al., 2019). Therefore, exploring perceptions of competence and hireability for more junior positions is important for future research, even more so because discrimination in the earlier stages of non-prototypical applicants’ careers would reduce their chances to apply later on for leadership positions at all. Once being established in a male-typed working environment women, be they White or Black, straight or lesbian, may no longer encounter discrimination based on their sexual orientation or race – as long as the person sifting their application is a woman (Cavaletto et al., 2019; Flabbi et al., 2019; Lucifero & Vigani, 2016; Theodoropoulos et al., 2022): In an exploratory analysis with participant gender, we found that men preferably wanted to keep heterosexual White applicants in the application pool compared to Black and/or lesbian applicants. Generally, men hold more anti-gay attitudes than women (Everly et al., 2015; Steffens & Wagner, 2004). A recent meta-analysis on gender-related discrimination found that men preferred male over female applicants for a male-typed job (A. J. Koch et al., 2016). This preference for choosing a prototypical candidate (a man for a male-typed job) might carry over to other social categories like race and sexual orientation and thus explain why male participants preferred the most prototypical applicant (a White heterosexual woman). The buffering effect of a non-normative sexual orientation found by Wilson et al. (2017) for gay men in the U.S. thus cannot be generalized to women in Germany.

Implications for theory. We used the Continuum Model as a theoretical basis for our research (Fiske & Neuberg, 1990). Accordingly, impression formation varies on a continuum from a rapid, purely category-based process (Stage 1) through confirmatory categorization (Stage 2) and re-categorization (Stage 3) to an entirely individualized process (Stage 4). It depends on the perceiver’s attentional resources (i. e., time and motivation), the relevance of the person to be assessed, and the results of previously run processes whether impression formation proceeds or stops after a specific stage. We posit that the obtained processing stage depends on interactions of social category memberships and individualizing information. Specifically, the re-categorization stage is of relevance because sub-grouping can occur with retained characteristics of the initial category and an integration of some individualizing characteristics that are unusual for the original category. If a person to be judged cannot be re-categorized based on their characteristics and other available information, Stage 4 involves the piecemeal integration of individualized information. This piecemeal integration could also come into play if no stereotypes exist. For example, it can be challenging to categorize a Black lesbian woman (see also Hutter & Wood, 2014). Piecemeal integration could mainly occur at intersections of several marginalized social categories, and when the respective stereotypes are interpreted as incongruent (Hutter & Wood, 2014; Steffens et al., 2016). Thus, piecemeal integration could be more likely for Black lesbian applicants. In contrast, black heterosexual women and White lesbian applicants could be re-categorized, while White heterosexual applicants could be processed in a more confirmatory manner. In other words, the more an applicant deviates from expectations, the more likely this person causes surprise (Hansen et al., 2017) and, therefore, is processed more individually.

However, according to our results, individuating information interacted only little with social group membership. The available individuating information (i. e., self-described competence levels) partially influenced the evaluation of applicants regarding inferred warmth, competence, hireability, and conservativeness. Thus, this could indicate that we either found piecemeal impression formation for all applicants, or at least, we did not find such piecemeal impression formation uniquely for non-prototypical applicants. The analysis of reaction times supports the idea that impressions of White heterosexual applicants were formed in a more stereotype-based way than those of other applicants, hinting towards quicker confirmatory categorization (Stage 2) of prototypical applicants when creating a first impression. In contrast, we found no evidence for the idea that information on Black lesbian applicants is processed particularly slowly, in a more individuated way (piecemeal integration) than on White lesbian or Black heterosexual applicants because of similar reaction times for these three social categories. Taken together, our findings provide only limited evidence for the Continuum Model.

Our findings add nuance to prior findings that lesbian applicants are perceived as more competent, warmer, and sometimes more hireable given individuating information (Niedlich et al., 2022). Lesbian applicants were descriptively judged to be higher in both competence and warmth than straight applicants, but only White lesbian applicants were seen as more progressive than straight applicants. The latter finding is consistent with other results showing that lesbian women are perceived as more progressive – but also more promiscuous, masculine, and confused (Jensen et al., 2022). It is possible then, that mixed positive and negative stereotypes of them were at play that levelled impressions of them. These results also underline the importance of an intersectional approach in research on discrimination because White lesbian and Black lesbian applicants were not perceived identically. The picture drawn by the Lack-of-fit model (Heilman, 2012) gets more nuanced when considering that not all women are alike. We contend that most of the research contributing to the formulation of the Lack-of-fit model has explicitly or implicitly reproduced a norm of straight White women when examining job-related gender discrimination. We found that, when explicitly stating an applicant’s sexual orientation and looking at different racialized groups, women can be seen as a good enough fit for a male-typed job. Our experiment thus underlines the importance of examining not only one social group membership while ignoring concurrent and co-constructed other relevant social identities.

Limitations and future research directions. Given that the residual variances were generally large compared to the explained variance across all multilevel models, a large part of the variance remains unexplained. We cannot rule out that this hints towards further, unknown influencing factors. Similarly, the lack of big differences in competence estimates across the competence conditions most likely influenced the fact that the cross-classified data structure could not be applied throughout as planned so that we partly had to fit a simpler hierarchical two-level structure. This could be due to the fact that our Level 2 regarding competence consisted of only four units, because simulation studies have shown that for a reliable assessment of random effects at least 10 units are preferable (Eid et al., 2015; Maas & Hox, 2004b, 2004a, 2005).

On a general note, we presented only female applicants because we wanted to account for the intersectional variety that exists within the group of women (Bowleg & Bauer, 2016), which will often be (implicitly) assumed to only comprise straight White women that are then compared to straight White men. Consequently, we could not test whether impressions of women were more positive, more negative, or similar to those obtained of male applicants. The decision to exclude male applicants was due both to practical considerations (i. e., an already large number of stimuli to maximize power for comparisons within the group of women) and to previous research where lesbian applicants were not discriminated as compared to heterosexual male applicants in male-typed jobs (Niedlich & Steffens, 2015). However, in the real world it is most likely that both female and male applicants would compete for a leadership position. Since women continue to be perceived as warmer than men (Ebert et al., 2014) and meta-analytic findings suggest discrimination in counter-stereotypical job contexts (Zschirnt & Ruedin, 2016) conducting an experiment with both male and female applicants could yield different results.

To make the high number of Black women applying more plausible, we emphasized in the cover story that all applicants were already part of the staff of an international hospital company with locations in several countries on all continents. This cover story has two potential caveats. First, working for an international company possibly increased competence impressions. Second, according to social identity theory (Tajfel & Turner, 1979) this statement could have created a sense of a global working community reducing the salience of other group memberships and thus activating a diversity norm. Even though this can be seen as a limitation of the experiment, it may be desired in an applied context.

We cannot rule out that sub-grouping of lesbian and heterosexual Black applicants occurred. As discussed by Ball and colleagues (2022) there are hints that Black targets in Germany are perceived to be foreigners sub-grouped either as African-American or as African, depending on their socioeconomic status. Because of the highly-qualified academic job-context, it is possible that the present applicants were perceived as having a high socioeconomic status and thus being African-American. This, in turn, could have signalled cultural similarity, thereby leading to no differences in the assessment of Black and White applicants for competence, warmth, and hireability (see, for example, Koopmans et al., 2019). Thus, for blue-collar jobs or with different names, more discrimination of Black applicants could be obtained (see Güngör & Biernat, 2008; Mansfield et al., 1991).

When interpreting some of the null effects of women’s sexual orientation on competence, warmth, and hireability judgments, it needs to be taken into account that our experiment included (smallish) photos of all applicants, matched for perceived masculinity (and attractiveness) because photos were needed to unobtrusively manipulate race. Thus, had we included masculinity ratings, effects of sexual orientation on perceived masculinity most probably would have been small (or nil). However, higher ascribed masculinity has been found to mediate higher competence impressions of lesbian than heterosexual women (Niedlich et al., 2014). Thus, in the absence of photos, differences in competence based on sexual orientation should more likely be obtained.

Furthermore, our experiment demonstrates that conducting multilevel analyses, which are needed to appropriately address a more complex data structure with clustering, resulted in overall smaller effects than analyses based on traditional ANOVAs. When using ANOVAs, White lesbians did appear particularly competent, for instance. If simpler designs are used, traditional ANOVAs can adequately model the data structure. Consequently, it is important to consider the actual data structure at hand to choose the strategy of data analysis that does neither over- nor underestimate effects. Thus, the data-analytic approach appears more crucial for the obtained null findings than presenting photos or not.

Finally, the composition of our sample needs to be discussed. On the one hand, a general-population sample representative of the German population regarding gender and age makes it less likely to detect statistically significant effects because of the large proportion of unexplained variance. Also, such a heterogeneous sample most likely contains both individuals with negative attitudes towards Black and lesbian women, who arrive at negative impressions of them, and individuals with positive attitudes arriving at positive impressions, such that positive and negative discrimination effects may have canceled each other out. In light of ongoing real-world discrimination (Flage, 2019), future research should collect more information on participants’ attitudes, which we omitted in order not to overburden participants because of our extended stimulus sampling, and thus assess moderators of ongoing discrimination. On the other hand, with our heterogeneous sample, the generalizability of the obtained findings is improved as compared to typical samples of young predominantly female university students with leftist attitudes.

Practice implications. In an online experiment with Black/White heterosexual/lesbian female applicants for a leadership position in a male-typed job we did not find evidence of negative or positive hiring discrimination on the basis of race, sexual orientation, or their intersection. This finding is encouraging in attesting to limits on discrimination based on social group membership, in line with a growing body of research (Fasoli & Hegarty, 2020; Livingston et al., 2012; Rosette et al., 2016; Wilson et al., 2017).

In such a counter-stereotypical job-context, female applicants were perceived to be quite competent independent of their self-description regarding different competence levels, and warmer, the more competence they expressed. White lesbian applicants were seen as more progressive than heterosexual applicants. Thus, a manipulation as subtle as presenting four adjectives (for self-description) varying in how much competence-warmth they convey, influenced perceivers’ first impression of job applicants when it comes to competence and warmth, and subtle information on applicants’ sexual orientation influenced applicants’ perceived progressiveness. This underlines the importance of even small portions of individuating information in countering stereotyping to a certain extent. Revealing/not hiding one’s sexual orientation from female supervisors (given that only male participants in our sample excluded Black or lesbian applicants more often from the application pool) may thus even have beneficial consequences for non-prototypical job incumbents (see also Petsko & Bodenhausen, 2019; Strinić et al., 2020).

However, we need to draw attention to the fact that we chose specific conditions to test our theoretically derived hypotheses: Because Black and/or lesbian women are stereotyped to be more masculine than heterosexual women, they could have an advantage regarding a male-typed leadership position. It is important to remember that at the same time, discrimination of Black and/or lesbian women is ongoing and to be expected regarding female-typed and lower-skilled jobs.

In a comprehensive online experiment, we did not find hiring discrimination of Black and/or lesbian women when applying for a male-typed leadership position. We tried hard to obtain generalizable findings on hiring discrimination by (i) extended stimulus sampling, (ii) obtaining a sample representative of the German population regarding gender and age, with prior work experience and some experience in personnel selection, (iii) including little diagnostic information on applicants’ qualifications by using only self-selected traits to signal individuating information on competence (and warmth), and (iv) using a male-typed job context in which discrimination of women is most likely. Individuating information in the form of self-descriptions conveying more or less competence and warmth did not substantially interact with race or sexual orientation in influencing competence and hireability ratings. However, it did influence hireability, warmth, and conservativeness ratings for all applicants regardless of race and sexual orientation.

This experiment enriches research by examining not only the intersections of gender with race and sexual orientation, but also the role of individuating information in hiring discrimination. We urge future research to rely on extended samples of stimuli to arrive at generalizable conclusions. Our null findings regarding competence and hireability, along with the overall high ratings of competence, hireability, and warmth thus indicate that there are circumstances under which social group memberships regarding sexual orientation and race affect impressions only negligibly and the situational context matters more. Whereas research is of course still needed that delineates the circumstances under which discrimination does in fact occur, experiments such as the present one are important to obtain a full picture avoiding publication bias.

We have no conflicts of interest, financial or otherwise.

All authors substantially contributed to conducting the underlying research and drafting this manuscript.

The current research was funded by the German Research Foundation [DFG STE 938/11-3].

We thank Nilanjana Dasgupta’s lab members for valuable comments on a previous version of this article and Nina Stillemunkes in helping with data collection.

All the stimuli, presentation materials, participant data, and analysis scripts can be found on this paper’s project page on OSF: https://osf.io/enakj/?view_only=b95e11fc4fc34ce5ab16f9dd0fb6be56.