1.1 The current study

Research on the attractiveness bias remains convoluted due to inconsistent results and context dependency (Lee et al., 2017; Nault et al., 2020). Research has started to explore how target sex or race might moderate the strength of the attractiveness bias. Initial research indicated that the attractiveness bias was not moderated by target sex—men and women would favor both attractive men and women (Eagly et al., 1991). However, more recent research indicates that this might not be the case. In research on job interviews, an attractiveness bias was observed for opposite-sex candidates, but participants discriminated against highly attractive same-sex candidates (Agthe et al., 2010, 2016). Similarly, in research on punishment, same-sex targets are punished more severely than opposite sex targets when the target is attractive, but this pattern is reversed when the target is less attractive (Li & Zhou, 2014; Mackelprang & Becker, 2017). This pattern of results is thought to have an evolutionary foundation, such that (at least for heterosexual persons) an attractive same-sex candidate is more likely to be perceived as a potential mating rival, inducing a threat response including emotions such as jealousy (Agthe et al., 2016).

In terms of race, some research has found that sex and race moderations do not predict hiring decisions (Marshall et al., 1998), whereas other research has found that an attractiveness bias (for opposite-sex candidates) emerges for same-race targets but not for other-race targets (Agthe et al., 2016). People from different ethnicities and cultures tend to agree on whom they consider attractive (Coetzee et al., 2014). However, the attractiveness bias might not be as strong when members of a different ethnicity are evaluated (Agthe et al., 2016). This is thought to occur because (1) members of other ethnic groups are less likely than same-ethnicity targets to be seen as potential mates and romantic rivals, (2) other-race targets are potentially less relevant for appearance-based social comparison because they appear less similar to the evaluator, and are therefore less likely to elicit self-other comparison, and (3) because people are more sensitive to mating-related cues in targets who share their own race (Agthe et al., 2016). This sensitivity to mating-related cues in same-race targets might result from socialization factors, such as greater family acceptance for a partner of the same race (Agthe et al., 2016).

The purpose of the current investigation was to further explore the moderating roles of sex and race in job recruitment. We explore both perception of candidate strength and likelihood to invite for interview as dependent variables. Perception of candidate strength should relate strongly to likelihood to invite for interview, but it can be useful to explore these as separate variables. That is, recruiters might (either consciously or unconsciously) use information other than perception of candidate strength when deciding whether or not to invite a candidate for interview. Therefore, it can be useful to explore likelihood to action (invitation) in addition to perception of strength. Moreover, being invited to interview is often the first barrier faced by job applicants, and therefore studying bias at this point is especially relevant to job recruitment. In addition, we include both a stronger applicant (strong resume) and a weaker applicant (weak resume) in the study design, to provide a benchmark variable for estimating the magnitude of bias effects. In an ideal world, where sex, race, and attractiveness are not factored into perceptions of job applicants, perceptions of candidate strength and likelihood to invite for interview should only differ across this objective resume strength variable.

Based on previous research (e.g., Agthe et al., 2016), we hypothesized a moderation effect for attractiveness and sex, such that participants (all women) would favor attractive male candidates over less attractive male candidates, but favor less attractive female candidates over more attractive female candidates (H₁). We also hypothesized a moderation effect for attractiveness and race such that participants (all White women) would favor attractive White candidates over less attractive White candidates, but not discriminate between attractive and less attractive Black candidates (H₂). Based on these two hypotheses, we also hypothesized a three-way interaction effect such that the tendency to favor attractive male candidates over less attractive male candidates (and favor less attractive female candidates over more attractive female candidates) would emerge for White candidates but not Black candidates (H₃). In addition, while attractiveness is valued across sexes it is most valued in potential mates (Buss & Schmitt, 1993). Therefore, the sexual orientation of participants might also be an important consideration in hiring decisions. In light of this, we explore data both with and excluding participants who identified their sexual orientation as nonheterosexual to check the robustness of results.

2.1 Participants

Sample size was determined using the statistical software package GPower 3.1 (Faul et al., 2007). For a four-way fixed effects analysis of variance with a medium effect size Cohen's f = 0.25, α error probability .05, and statistical power 1–β error probability .80, the total sample size required is 320. A medium effect size was targeted based on effect sizes identified in previous research exploring moderators of the attractiveness bias (e.g., Allen et al., 2019; Swami et al., 2017). Due to the potential confounding factors of participant sex and race (see Agthe et al., 2016), only White women were included as participants in the current study. All participants were undergraduate psychology students recruited from a single university. Research has supported the validity of using student participants in place of professional evaluators in vignette research, with no significant differences observed between ratings (Hosoda et al., 2003). In total, 338 White women (age range = 17 to 57 years; M_age = 20.94, SD = 5.65) agreed to take part in the research. In total, 270 reported their sexual orientation as heterosexual or straight, and 68 reported their sexual orientation as lesbian, bisexual, or other.

2.2 Materials

2.3 Procedure

Before conducting the study, ethical clearance was obtained from a university research ethics committee. Participants were recruited from a single university and received course credit in exchange for their participation. Participants provided digital informed consent and provided information on sex, age, ethnicity, and sexual orientation. Only participants that reported their sex as female and ethnicity as White were included in the study. Participants were informed that the aim of the study was to investigate factors influencing perceived strengths and hireability of job applications, and were randomly allocated 1 of 16 vignettes. They were instructed to imagine they were in a situation whether they had 20 job applications and needed to select 5 to be interviewed. They were then provided the job advertisement and a single application from 1 of the 16 vignettes and were asked to rate the applicants' strength and the likelihood of inviting the candidate for an interview. Once this was completed, participants were informed about the true purpose of the study. Participants completed the study online and unsupervised.

2.4 Analyses

Two 2 (target sex: female, male) x 2 (application strength: strong, weak) x 2 (target race: Black, White) x 2 (target attractiveness: attractive, less attractive) between-subjects analysis of variance were computed on the two dependent variables of applicant strength and likelihood of interview. Homogeneity of variance was assessed using Levene's test of equality of error variances, normality was checked through observations of data skewness (<±2.00 indicates a normal distribution), and potential outliers were identified through z-scores. One potential outlier was identified on the dependent variable likelihood to interview (z = 3.08). Analyses were run with and with and excluding this outlier to check the robustness of results (a sensitivity analysis). Homogeneity of variance was violated for both strength of applicant (p = .025) and likelihood to interview (p = .015). However, analysis of variance (ANOVA) is robust to violations of homogeneity of variance in larger sample sizes where minor departures from normality can often appear statistically significant (Blanca et al., 2017). Significant interaction effects were followed up using independent samples t-tests.

As opposite sex candidates are predicted to show a greater attractiveness bias (due to sexual attraction; see Agthe et al., 2016), this effect might be contingent on participant sexual orientation. Therefore, all analyses were re-run on a subsample of participants that described their sexual orientation as heterosexual or straight (N = 270). The subsample included 270 White women (age range = 17 to 57 years; M_age = 20.89, SD = 5.77). In the subsample, all data appeared normally distributed and z-scores revealed no potential outliers. Homogeneity of variance was confirmed for both strength of applicant and likelihood to interview. In terms of interpretation, an effect size d = 0.20 (η² = 0.01) was considered small, an effect size d = 0.41 (η² = 0.04) was considered medium, and an effect size d = 0.63 (η² = 0.09) was considered large.

3.1 Perceived strength of applicant

The ANOVA revealed a significant main effect of target race, F(1, 337) = 4.16, p = .042, $$\eta $$² = 0.01, with Black targets (M = 4.62, SE = 0.09) receiving higher ratings than White targets (M = 4.37, SE = 0.09). There was also a significant main effect of application strength, F(1, 337) = 100.33, p < .001, $$\eta $$² = 0.24, such that strong applications were rated higher (M = 5.12, SE = 0.09) than weak applications (M = 3.87, SE = 0.09). There was no significant main effect found of target attractiveness, F(1, 337) = 8.12, p = .368, $$\eta $$² < 0.01, or sex, F(1, 337) = 0.92, p = .338, $$\eta $$² < 0.01. In terms of moderation effects, there were no significant two-way interaction effects for target attractiveness and target race, F(1, 337) = 2.23, p = .137, $$\eta $$² = 0.01, target attractiveness and target sex F(1, 337) = 3.42, p = .065, $$\eta $$² = 0.01, or target sex and target race F(1, 337) = 0.59, p = .444, $$\eta $$² < 0.01. The three-way interaction effect for target sex, race and attractiveness was also nonsignificant, F(1, 337) = 2.21, p = .138, $$\eta $$² = 0.01, and there were no significant interaction effects involving target application strength. These effects are depicted in Figures 1 and 2.

For the subsample of heterosexual participants (n = 270), the ANOVA again revealed a significant main effect for application strength, F(1, 269) = 76.80, p < .001, $$\eta $$² = 0.23, with strong applications rated higher (M = 5.08, SE = 0.10) than weak applications (M = 3.87, SE = 0.10). There were no significant main effects found for target attractiveness, F(1, 269) = 0.72, p = .397, $$\eta $$² < 0.01, target race, F(1, 269) = 1.44, p = .232, $$\eta $$² = 0.01, or target sex, F(1, 269)= = 1.08, p = .300, $$\eta $$² < 0.01. For moderation effects, there were no significant two-way interactions observed for target attractiveness and target race, F(1, 269) = 2.33, p = .128, $$\eta $$² = 0.01, target attractiveness and target sex, F(1, 269) = 2.11, p = .148, $$\eta $$² = 0.01, or target sex and target race, F(1, 269) = 0.99, p = .321, $$\eta $$² = 0.01. The three-way interaction effect for target sex, race and attractiveness was also nonsignificant, F(1, 269) = 0.38, p = .539, $$\eta $$² < 0.01. However, there was a significant three-way interaction for target sex, target race and application strength, F(1, 269)= = 4.93, p = .027, $$\eta $$² = 0.02. However, analysis of simple effects showed that the interaction was nonsignificant on all levels of the independent variables indicating that this three-way interaction is probably not meaningful.

3.2 Likelihood to interview

For likelihood to interview, there was a significant main effect for target race, F(1, 337) = 9.25, p = .003, $$\eta $$² = 0.03, with Black targets (M = 4.82, SE = 0.10) being more likely to be invited to interview than White targets (M = 4.38, SE = 0.10). There was also a significant main effect for application strength, F(1, 337) = 93.89, p < .001, $$\eta $$² = 0.23, with strong applications (M = 5.30, SE = 0.10) being more likely to be invited to interview than weak applications (M = 3.90, SE = 0.10). There was no significant main effects for target attractiveness, F(1, 337) = 1.93, p = .166, $$\eta $$² = 0.01, or target sex, F(1, 337) = 3.18, p = .075, $$\eta $$² = 0.01. For moderation effects, there was a significant two-way interaction for target attractiveness and target race, F(1, 337) = 4.43, p = .036, $$\eta $$² = 0.01. This effect was such that attractive White candidates were more likely to be invited to interview (M = 4.63, SE = 0.14), than less attractive White candidates (M = 4.13, SE = 0.15), t(167) = 2.32, p = .022, d = 0.36, but there was no significant difference between attractive Black candidates (M = 4.77, SE = 0.14) and less attractive Black candidates (M = 4.87, SE = 0.15), t(167) = 0.45, p = .654, d = 0.07.

There were no significant two-way interaction effects for target attractiveness and target sex, F(1, 337) = 2.99, p = .085, $$\eta $$² = 0.01, or target sex and target race, F(1, 337) = 2.12, p = .147, $$\eta $$² = 0.01. There was also no significant three-way interaction between target attractiveness, target sex and target race, F(1, 337) = 0.00, p = .963, $$\eta $$² < 0.01. However, there was a significant three-way interaction for target race, target sex and application strength, F(1, 337) = 8.36, p = .004, $$\eta $$² = 0.03. Analysis of simple effects showed that this effect was such that, for stronger applications only, Black candidates were more likely to be invited to interview than White candidates, but only for men, t(79) = 2.65, p = .010, d = 0.60. For weaker applications, Black candidates were more likely to be invited to interview than White candidates, but only for women, t(86) = 3.47, p = .001, d = 0.75. These findings are depicted in Figures 3 and 4, and findings remained unchanged with one potential multivariate outlier removed from the sample.

For the subsample of heterosexual participants (n = 270), the ANOVA revealed a significant main effect for target sex, F(1, 269) = 4.33, p = .038, $$\eta $$² = 0.02, with female candidates (M = 4.75, SE = 0.11) more likely to be invited to interview than male candidates (M = 4.41, SE = 0.12). There was also a significant main effect for application strength, F(1, 269) = 74.17, p < .001, $$\eta $$² = 0.23, with strong applications (M = 5.27, SE = 0.11) more likely to be invited to interview than weak applications (M = 3.89, SE = 0.11). There was no significant main effect found of target attractiveness, F(1, 269) = 0.71, p = .401, $$\eta $$² < 0.01, or target race, F(1, 269) = 3.49, p = .063, $$\eta $$² = 0.01. For moderation effects, there was a significant two-way interaction effect for target sex and application strength, F(1, 269)= 3.95, p = .048, $$\eta $$² = 0.02. This effect was such that female candidates (M = 4.22, SE = 0.15) were more likely to be invited to interview than male candidates (M = 3.56, SE = 0.16) when the application was weak, t(135) = 2.65, p = .009, d = 0.46. However there was no difference between female candidates (M = 5.28, SE = 0.16) and male candidates (M = 5.26, SE = 0.16) when the application was strong, t(131) = 0.35, p = .728, d = 0.06. There were no significant two-way interaction effects for target attractiveness and target race, F(1, 269) = 3.34, p = .069, $$\eta $$² = 0.01, target attractiveness and target sex, F(1, 269) = 1.05, p = .307, $$\eta $$² < 0.01, or target sex and target race, F(1, 269)=;= 2.48, p = .117, $$\eta $$² = 0.10. There was no significant three-way interaction effect for target attractiveness, race and sex, F(1, 269) = 0.26, p = .610, $$\eta $$² < 0.01. However, there was a significant three-way interaction effect for race, sex and application strength, F(1, 269) = 4.93, p = .027, $$\eta $$² = 0.02. This effect was such that, for stronger applications, Black candidates were more likely to be invited to interview than White candidates, but only for men, t(64) = 2.21, p = .031, d = 0.55. For weaker applications, Black candidates were more likely to be invited to interview than White candidates, but only for women, t(70) = 3.09, p = .003, d = 0.74.