What is in your résumé? The effects of multiple social categories in résumé screening

Abstract

Purpose

Numerous studies have shown that minority workers are disadvantaged in the labour market due to stereotypes and discrimination. However, published research on résumé screening has overlooked the effects of multiple social categorisations pertaining to candidates' gender, education and origin. This study addresses this gap and examines whether the gender, the level of education and the national origin cues mentioned in the résumé affect the perceived employability of candidates.

Design/methodology/approach

This study employs an experimental between-subjects factorial design in that 12 résumés varying in gender, education and national origin were rated by 373 Portuguese working adults.

Findings

The results documented a gender premium as women were favoured in interpersonal and job skills but not in job suitability, and an education premium, since higher educated candidates were preferred despite their gender and origin. No meaningful interactions for gender × education × national origin were observed, which suggests that ingroup favouritism and outgroup discrimination in résumé screening can be averted.

Originality/value

The findings endorse a multidimensional view of perceived employability by investigating candidates' skills and job suitability from the viewpoint of the decision-makers, which extends our understanding of résumé-screening discrimination. This is critical to prevent hiring discrimination at an earlier career stage, which can increase youth employment and enhance the integration in the labour market of local minorities such as women, inexperienced workers and second-generation immigrants.

Citation

Pinto, L.H., Portugal, R. and Viana, P. (2023), "What is in your résumé? The effects of multiple social categories in résumé screening", Personnel Review, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/PR-04-2023-0290

Introduction

National origin, along with other demographics like age, gender and education, are commonly used categorisation schemes (Bauer and Hannover, 2020; Moore-Berg and Karpinski, 2019). The tendency to categorise and rate others as members of social groups rather than individuals are stronger when the information available is limited, as is the case with résumé screening. Explicit and implicit stereotypes associated with social group memberships are likely to influence recruiters' impressions of job applicants, which can determine the chances of employment.

Traditionally, people feel more attracted to and prefer others like them, and this ingroup favouritism (Tajfel and Turner, 1986) has been empirically confirmed in the hiring setting. In addition, discrimination toward outgroup minorities, including second-generation immigrants (Midtbøen, 2016), has been well documented (for a review, see Zschirnt and Ruedin, 2016). Immigrant workers are less frequently selected for job interviews (e.g. Derous et al., 2016) and promotions (e.g. Bastida and Moscoso, 2015), and they also experience explicit and implicit discrimination in the workplace (Hebl et al., 2020). Recruiters who are ethnically less diverse, less experienced and high in social dominance orientation (SDO) are also more likely of hiring discrimination (Derous, 2017).

Hiring research has mainly focused on the discrimination of low-status minority groups, such as the Black and Latinos in America and the poorest European and non-European immigrants, including Arabs, Turks and Chinese (Ford and Mellon, 2020). Prior studies have confirmed a hierarchy of minorities in Europe: Arabs, Chinese and people from the Middle East are generally more discriminated against than other European migrants (Derous et al., 2012; Ford and Mellon, 2020), and there is a preference for highly skilled immigrants. However, this choice of distant outgroups leads to overestimating the extent of discrimination, as it frequently disregards what happens when origin cues combine with other attributes, such as gender and local education, potentially decreasing outgroup discrimination. In other words, research on hiring discrimination has rarely examined the nuanced stereotypes of candidates belonging to multiple social categories simultaneously and has not considered the perceived similarity of these candidates to the local majority. Research comparing the perceived employability of second-generation immigrants from low vs high-privileged national origins remains scarce (for a recent exception, see Ford and Mellon, 2020), and most studies failed to test the effects of recruiters' characteristics on résumé screening (for exceptions, see Derous, 2017; Pinto and He, 2019).

These shortcomings are particularly important in studying hiring decisions because in multiple categorisation settings, high-status others (e.g. high in education, job skills or immigration rank), including second-generation immigrants, can be perceived as “ingroup-like” and similar to the rater even when they do not belong to the same group (Grigoryan, 2020). Multiple social cues embedded in the résumé can induce nuanced impressions of candidates' overall similarity with the rater (Grigoryan, 2020) and higher interpersonal attraction, which may influence hiring decisions (Derous and Ryan, 2019).

This study tests these assumptions by building upon the stereotype content model (Cuddy et al., 2011) and the similarity-attraction theory (Byrne, 1971; Montoya and Horton, 2013) to examine the effects that multiple social categories included in the résumé can have on the perceived similarity and employability of the candidates. The study employs an experimental between-subjects factorial design in that 12 résumés operationalise the research variables: gender (male × female), education (high × vocational), and national origin (Portuguese: local majority × British: local European minority × Chinese: local non-European minority). Two disparate minorities varying in immigration rank were chosen to test possible employability differences between the two outgroups. The perceived similarity of the candidate (with the rater) and the employability of the candidates were then reported, and this was later measured by a set of skills relevant to perform the same entry-level accounting clerk job (Del Baldo et al., 2019; Hall et al., 2019), such as interpersonal and job skills, and job suitability. This job was selected because it is gender-neutral and can be performed by a recent graduate or someone with vocational training who has little or no work experience (Cooper and Robson, 2006), as confirmed by local statistics on gender representation and level of education of clerical workers (National Institute of Statistics, 2023a, b).

The contributions of this study are manifold. Firstly, the study addresses earlier calls to contextualise the effects of multiple social categorisations in résumé screening (Derous et al., 2012; Tsai et al., 2011) and hiring discrimination (Derous and Pepermans, 2019) to understand its implications for workplace diversity (Fletcher and Beauregard, 2022). By comparing the résumé screening of locals (Portuguese) with two other national origins (British × Chinese) also varying in gender and local education, one extends our understanding of hiring discrimination in a European context, including the extent of discrimination towards second-generation immigrants. Secondly, the variables of interest are the perceived similarity (with the rater), the job skills and the employability of the candidates. This is an important distinction to determine if applicants are discriminated against because they are seen as (dis)similar or because they are not skilled, job suited or both. In addition, this approach endorses a multidimensional view of perceived employability (Vanhercke et al., 2014) by investigating candidates' skills and job suitability from the viewpoint of the decision-makers, which extends our understanding of the effects of résumé-screening discrimination. Thirdly, this study uses résumés of young and qualified applicants for whom formal education is fundamental, and the résumé is still a key instrument in entering the labour market (Cole et al., 2007, 2009). Understanding how the multiple social categories present in the résumé intersect and influence the employability of these candidates is then critical to prevent hiring discrimination at an earlier career stage, which can then contribute to increasing youth employment and enhance the integration in the labour market of local minorities such as women, inexperienced workers and second-generation immigrants. This study aims not only to help theory integration but also to offer important insights for European policymakers in realising how to prevent discrimination early on and during résumé screening, which is necessary to promote workforce diversity. Finally, this study offers a methodological addition to the field experiments most commonly used in résumé screening that employ relative call-backs and odd ratios call-backs. Instead, this study uses a survey experiment that targets Portuguese working adults from diverse organisations rather than students or just recruiters since they are active resumé screeners in small and medium companies (Holden and Jameson, 2002), which are dominant in Europe and contribute to more than 66% of total EU employment (Rotar et al., 2019). Furthermore, respondents' demographics were included in the analysis, which addresses recent calls (e.g. Derous and Ryan, 2019; Derous et al., 2012) to examine how raters' characteristics and attributes can influence résumé screening and recruitment decisions.

The remaining sections of this paper describe the relevant literature, methodology and results, discuss the study limitations and the theoretical and managerial implications, and future research suggestions. Concluding remarks are then presented in the final section.

Stereotypes in résumé screening

According to the stereotype content model (SCM), group stereotypes are shared beliefs associated with an established social category that vary along two core dimensions (Cuddy et al., 2008, 2011; Fiske, 2012): perceived competence (i.e. skills, capacity) and perceived warmth (i.e. friendliness, trustworthiness). Variations in perceived warmth and competence resulting from interpersonal and intergroup interactions can lead to negative and positive stereotypes (Fiske et al., 2002). These two dimensions are then used to classify the members of the ingroup, which are usually high in competence and high in warmth, and rival group members, which are low in warmth and competence or nuanced (i.e. high/moderate in one dimension and medium/low in the other). Ingroup members are usually warm and competent, eliciting admiration and respect (Fiske et al., 2002). Meanwhile, derogated outgroup members, such as immigrants and other minorities, are seen as hostile and incompetent (Cuddy et al., 2009).

Across cultures, age, gender and national origin stereotypes have shown variations in intensity but not patterns in that sexism and ageism have been consistently reported in distinct settings. However, national stereotypes are more likely to vary with historical and national circumstances and migration patterns (Fiske, 2012), so studying stereotypes is best examined in context (Cuddy et al., 2009). Given that the résumé information is limited and based on candidates' attributes, such as gender, education and origin (among others), these multiple social categories are likely to produce nuanced stereotypes (Cuddy et al., 2009), often pejorative for minorities. Still, they can also be positive (Czopp et al., 2015). This interplay and its stereotypic effects on résumé screening are the focus of this research and are discussed in more depth below.

Gender premium

Gender stereotypes remain common across countries, although with some cultural variations (Fiske, 2017). According to the SCM, women are usually perceived as warm and socially oriented, and men are agentic and competent (Ellemers, 2018; Fiske, 2012). Zschirnt and Ruedin's (2016) meta-analysis on hiring discrimination found that while non-national women scored better on job suitability, the gender differences were not statistically significant. Hence, they found no systematic gender preference on a large scale. Some prior studies have not found gender differences in graduates' employability (Pinto and Ramalheira, 2017), except for interpersonal skills (Pinto and Pereira, 2019), for which female graduates were better regarded.

According to the SCM, women are commonly stereotyped as warmer and superior in interpersonal skills, although not necessarily more competent than men (Fiske, 2017). Hence, women may score higher in the “soft” skills (Andrews and Higson, 2008) that are stereotypically feminine (Ellemers, 2018), such as interpersonal skills, even for jobs that are gender-neutral and that do not specifically require these attributes. This stereotypical assessment may prevail despite applicants' education and origin. For the attributes that depend on candidates' credentials and fit the job requirements, such as job skills and job suitability, gender preferences are not expected (Zschirnt and Ruedin, 2016). Therefore, the following hypothesis is proposed:

H1.

Gender premium: Higher ratings of (a) interpersonal skills are expected for female candidates in comparison with male candidates (regardless of the level of education and national origin), whereas no gender differences are expected for the ratings of (b) job skills and (c) job suitability.

Education premium

The positive influence of higher education (HE) on job prospects and potential earnings have long been documented, holding even when workers are more skilled than the job requires (Daly et al., 2000; Ford and Mellon, 2020). Higher education generally signals more advanced cognitive skills (Piopiunik et al., 2020) and refined interpersonal and employability skills (Scott et al., 2019; Succi and Canovi, 2019), which employers value (Green and Henseke, 2021).

This preference for higher-educated (HE) candidates is consistent with the predictions of the SCM. While a medical degree is a strict prerequisite for being a surgeon, a management degree is not necessary to perform an entry-level accounting clerk job (Cooper and Robson, 2006; Green and Henseke, 2021). However, the skills requirements have been rising, so 23% of all local clerical workers already have high education (National Institute of Statistics, 2023a). Therefore, a preference for HE candidates based on their superior competence should signal an education premium. This preference might apply to the assessment of national and non-national applicants (Ford and Mellon, 2020), especially when the candidates born abroad report a local degree (McGuinness and Byrne, 2015), and thereby, are more likely to be seen as “ingroup like”. Conversely, the raters will consider candidates who only have vocational training less reputable and, therefore, dissimilar to themselves even when the raters personally do not have high education. These candidates will score lower on job skills and employability. The following hypothesis is then advanced:

H2.

Education premium: Higher ratings of (a) perceived similarity (with the rater), (b) interpersonal skills, (c) job skills and (d) job suitability are expected for the candidates with high education in comparison to the candidates with vocational education (regardless the gender and national origin).

National origin premium

In the European context, several field experiments provide contradictory evidence on hiring discrimination against local minorities. Some studies found discrimination (Blommaert et al., 2014; Derous et al., 2012, 2016) also against second-generation immigrants (Midtbøen, 2016), while other studies employing less stereotyped minority groups found smaller (McGinnity and Lunn, 2011) or no hiring differences (Alecu, 2019; Derous et al., 2009). However, Zschirnt and Ruedin's (2016) meta-analysis found discrimination even towards non-local candidates who were schooled locally (i.e. second-generation immigrants). Although some recent studies documented low perceived discrimination experienced by Cape Verdean (Neto et al., 2022) and Indian immigrants (Neto and Neto, 2023) living in Portugal, hiring discrimination against outgroup candidates remains pervasive in most contexts. From a social identity perspective (Tajfel and Turner, 1986), local candidates are expectedly preferred (i.e. ingroup favouritism) over equally qualified minority candidates who would be derogated (i.e. outgroup discrimination). Therefore, the following hypothesis can be advanced:

H3.

Ingroup favouritism: The highest ratings of (a) perceived similarity (to the rater), (b) interpersonal skills, (c) job skills and (d) job suitability are expected for the Portuguese candidates (regardless of gender and education).

Interaction effects

In testing the SCM in Europe, Cuddy et al. (2009) found several country stereotypes based on competence and warm differences, while some other studies found that national origin has a negative impact only in certain contexts (McGinnity and Lunn, 2011; Alecu, 2019; Derous et al., 2009), and for less prestigious ingroup members (Lewis and Sherman, 2003) and minorities (Ford and Mellon, 2020).

This means that ingroup favouritism and outgroup discrimination can vary, and ingroup derogation and outgroup favouritism can be observed. In testing these propositions, Dietz et al. (2015) failed to find support for ingroup favouritism/outgroup discrimination as they found no significant differences regarding the job adequacy of less qualified locals and immigrants. They interpreted this result as a form of “conditional ingroup bias” aimed to protect the ingroup from the least skilled members (Dietz et al., 2015), which confirms the possibility of ingroup denigration (Lewis and Sherman, 2003). Likewise, and for low cognitive demanding jobs, Derous and Pepermans (2019) have not found outgroup discrimination as there were no significant differences in the job suitability of natives (Belgian men/women) and foreigners (Maghreb-Arab men/women). In high cognitive-demanding jobs, they even observed outgroup favouritism through a preference for the Maghreb/Arab male applicants, which supported earlier findings (Derous et al., 2009). Also, in testing hiring discrimination towards highly qualified candidates in Norway, no bias or overall trust variation was observed regarding local Norwegian vs Pakistani doctors (Alecu, 2019).

In sum, these findings suggest that: (1) ingroup derogation can be observed towards low-status/less educated ingroup candidates and (2) outgroup favouritism can emerge towards high-status/highly educated outgroup members. These predictions are consistent with the SCM (Cuddy et al., 2009). For example, Cuddy et al. (2009) found that the UK was stereotyped in the highest-competence/lowest-warmth cluster, while Portugal clustered in the low-competence/high-warmth quadrant. They found that the Portuguese rated themselves significantly higher than other nations rated them on warmth but not in competence. In other words, people can ensure a positive differentiation by favouring the in-group dimension relevant to their identity while still recognising the outgroup on the other dimension (Cuddy et al., 2009; Lewis and Sherman, 2003). Consistently, British immigrants in Portugal are seen by locals as “upper-class” and highly competent (Oliveira and Gomes, 2019), and when compared to other local minorities, like the Chinese, British immigrants are generally more qualified, perform higher intellectual and scientific positions and are better paid (Oliveira and Gomes, 2019).

In applying the SCM to the Portuguese context, it is reasonable to expect a derogation of the non-educated members of the ingroup, who might score lower than the highly educated minority applicants (ingroup derogation), especially the HE British candidates. Locals see British immigrants in Portugal as “upper-class” and highly competent (Gaspar and Rodrigues, 2021). Local statistics confirm that they are highly qualified, perform intellectual and scientific positions and are better paid than locals (Oliveira and Gomes, 2019). Regarding the Chinese, they belong to a non-European minority frequently discriminated against in the labour market (Ford and Mellon, 2020; Gaspar and Rodrigues, 2021) and seen as less competent and warm (Cuddy et al., 2009), so ingroup derogation vis-à-vis the Chinese candidates is not expected to be noteworthy. Given that interpersonal skills (i.e. warmth) are relevant to the ingroup (Cuddy et al., 2009) but not to the outgroups, one expects no derogation for this attribute on the less-educated Portuguese candidates. Therefore, the following hypotheses are offered:

H4.

Ingroup derogation: Lower ratings of (a) perceived similarity (to the rater), (b) job skills and (c) job suitability are expected for the Portuguese candidates without HE, in comparison with the HE British candidates but not the HE Chinese; whereas (d) no differences are expected for interpersonal skills.

Finally, a premium towards the HE British candidates (i.e. outgroup favouritism) might also be observed in high-status/high-education conditions, especially in comparison with the other minority. Although the Chinese are often stereotyped as industrious (see Guo, 2022, for a discussion on Chinese Diasporas), they remain discriminated against in Europe (Ford and Mellon, 2020) and locally (Gaspar and Rodrigues, 2021; Oliveira and Gomes, 2019; Pinto et al., 2023), including the Chinese international students (França et al., 2022); so, HE British candidates can be seen as more competent than the HE Chinese candidates (Ford and Mellon, 2020), scoring higher in job skills and job suitability. However, the extent of outgroup favouritism for the HE British candidates is not expected to overrun the ingroup favouritism towards the HE locals (Alecu, 2019), so no differences are expected between the equally qualified Portuguese and British candidates. The following hypothesis is then presented:

H5.

Outgroup favouritism: Higher ratings of (a) perceived similarity, (b) job skills, and (c) job suitability is expected for the British candidates with higher education in comparison to the equally qualified Chinese candidates.

Methods and design

Procedure and sample

This study employed an experimental between-subjects factorial design since it is a usual and adequate approach (Wulff and Villadsen, 2020) that permits the control of recruiter-based effects, which is not feasible through correspondence audit studies. Three independent variables were manipulated: applicants' gender (male × female), education (vocational × HE) and national origin (Portuguese: local majority × British: European minority × Chinese: non-European minority). An online survey was employed to collect the data and target Portuguese working adults working in diverse organisations. These subjects were intentionally sought because they have work experience and are commonly involved in recruiting new employees (Holden and Jameson, 2002). Participation was voluntary and unpaid, and potential respondents were approached through social media (e.g. LinkedIn and Facebook) and researchers' networks in the local business community.

By consenting to participate in the study, respondents were randomly assigned to read one of the twelve fictitious résumés that portrayed a candidate for an entry-level accounting clerk job. The reasons for choosing this job are fourfold. Firstly, workers with and without higher education perform clerical jobs in Portugal. Given the local high unemployment rate of young graduates–in 2022, it was 9.4% for all people under 34 years old and 6.8% for graduates (National Institute of Statistics, 2023a) – it is not uncommon to find graduates performing this starting job. Secondly, accounting clerk jobs are receiving recruiting attention in the local market, given the establishment of many Shared Service Centres specialised in providing international accounting and administrative services (Figueiredo and Pinto, 2021). Thirdly, this job is gender neutral: the total number of Portuguese residents having a clerical job (in accounting, finance and insurance) in 2021 amounted to a total of 17.688 workers, of which 51% were male (i.e. 9.603) and 49% were female (i.e. 8.625) (National Institute of Statistics, 2023b). Finally, this job is not locally associated with any specific foreign group, as with other jobs, such as cleaning or retail, which are more represented, respectively, among immigrants from Portuguese-speaking countries and Chinese (National Institute of Statistics, 2023c).

Following, a set of questions regarding the candidate and related to the dependent variables were asked. In the end, the participants were asked to report personal information to characterise the sample (e.g. gender, age, education, occupation) and control for potential raters' biases (e.g. international residing and recruiting experience), including information about their place of birth, ethnicity and citizenship, to ascertain they all belong to the same national/ethnic group.

All respondents were white Europeans and Portuguese (i.e. born in Portugal, having Portuguese citizenship and belonging to the local majority of white Europeans). In total, 444 persons answered the questionnaire. After excluding 26 replies from non-locals and non-working respondents (i.e. students, unemployed and retired) and 44 from respondents who wrongly answered the manipulation check, the final sample consisted of 373 Portuguese working adults, as presented in Table 1.

The average age was 31.60 years old, and 56.3% of the respondents were female. Respondents were highly qualified (only 10.8% did not have a university degree) and were exposed to cultural diversity: 124 individuals (33.8%) had lived abroad, and 214 (58.5%) had family living abroad. Respondents were employed in medium (37.2%) to large (45%) private organisations (83.8%), with an international workforce (63.9%). Most respondents had professional and specialised occupations (62.8%), including senior managerial roles (24.4%) and recruiting experience (50.4%).

Stimulus materials

To standardise the experimental conditions, a basic fictitious résumé was developed drawing upon the versions of Pinto and Ramalheira (2017) for a Portuguese business graduate. Each single-page résumé contained descriptive information about a 21-year-old candidate applying for an entry-level accounting clerk job. The layout of the résumé, as well as the remaining content (e.g. personal data, professional aims, education and additional training), reflect the usual résumé content sought by recruiters (Brown and Campion, 1994; Hiemstra et al., 2013). No photo was inserted to control for a potential attractiveness bias (Apers and Derous, 2017). Likewise, no prior work experience was mentioned, so the résumés portrayed a young candidate entering the labour market. From the basic male/female versions, two other variations were employed for the level of education and national origin. Education was manipulated by referencing a vocational diploma versus a bachelor's degree granted by local public institutions, respectively, for the conditions of no-HE and HE. National origin was manipulated through the use of three distinct surnames, as per similar approaches by Cotton et al. (2008) and Derous and Ryan (2012): a popular Portuguese surname for the candidates of the majority group (i.e. Martins Antunes), and common surnames to Chinese (i.e. Wei Wang) and British (i.e. Taylor Smith) candidates. To prevent perceptions of an illegal migratory status, all resumés reported local residence and education (i.e. academic credentials granted by local institutions) and proficiency in the native and local language. This material is available in Appendix.

Given that the content of the résumés could be misattributed, especially when using non-national surnames (Cotton et al., 2008), all résumés were printed and pre-tested. One focus group with Portuguese human resource professionals were used to determine the realism/credibility of the resumés, the relevance/adequacy of educational qualifications and the believability/authenticity of the surnames as cues for different national origins. As a result, a few minor graphic amendments were made, so the 12 résumés were considered comparable, equivalent and illustrative of candidates from different national origins. In addition, HR professionals corroborated the adequacy of the résumés to an entry-level accounting clerk job, confirming that this job is gender and ethnic-neutral.

Measures

Perceived similarity

This is a one-item scale based on Deprez-Sims and Morris (2010) that asked: “In your opinion, how similar is this candidate to yourself?” Response options ranged from 1 (very low similarity) to 5 (very high similarity). This choice follows previous studies (e.g. Grigoryan, 2020) that have successfully used similar single-item measures of perceived similarity related to multiple social categorisations. By using a generic measure without a clear indication of the kind of similarity, the raters had to look for (e.g. gender, education, national origin), one aimed to control for this variable while minimising demand effects, potentially introducing bias and distorting the results (Lonati et al., 2018).

Interpersonal skills

Measures candidates' warmth and friendliness at work through a 5-item scale that assesses how well the candidate may relate to others (Evers et al., 1998). A 5-point response Likert scale was used (1 = very low competence/ability level; 5 = very high competence/ability level), and sample items included: “This candidate is capable of understanding other's needs”, and “This candidate is capable of working well with others (superiors, subordinates, and peers)”. The original reliability for this scale was 0.80 (Evers et al., 1998), whereas in this study was 0.884, which is considered very good and supports its use.

Job skills

Measures the perceptions of applicant's competence to perform the target job through a 5-item scale adapted from Evers et al. (1998). This variable measures the degree to which the job applicant is competent to: “Keeping up-to-date on developments in the (professional) field”, “Manage and supervise several tasks simultaneously” and “Is able to set priorities”. Items were answered using a 5-point Likert scale (1 = very low competence/ability level; 5 = very high competence/ability level). The Cronbach alpha for this scale was 0.869, higher than the original score (α = 0.83) from Evers et al. (1998).

Job suitability

Measures the job suitability of each candidate through a 5-item scale, adapted from McElroy et al. (2014). Sample items included: “This candidate is a good match for the position”; “This person has a good chance of making a ‘short list’ of candidates for this position”; “I would not hire this person for this position” (reverse coded). Response options ranged from 1 (totally disagree) to 7 (totally agree). In this study, the scale's reliability was 0.897, which compares well with earlier findings (0.86, according to Pinto and Ramalheira, 2017).

Other measures

To control for other potential respondents' biases associated with comparisons with the target (e.g. having lived abroad and/or having an international background), all subjects were asked about their age, gender, education, national origin, citizenship, ethnicity, previous residence abroad, family living abroad and recruiting experience. National origin and citizenship (Portuguese), ethnicity (white European) and work status (employed) were used to rule out non-national respondents and choose the final sample. Age was computed in years. Gender was dummy-coded (0 = Female; 1 = Male) and Education (0 = Less than higher education; 1 = higher education). Recruiting experience was dummy-coded, as previous residence abroad and family living abroad (0 = No; 1 = Yes).

Manipulation checks

Respondents indicated whether or not each of the eight pieces of information was present in the résumé they read and how sure they were about the presence of that content, using a 7-point scale, ranging from 1 = very unsure to 7 = very sure. Two items focused on the gender of the applicant (male/female), two other items referred to the education reported in the résumés (vocational diploma/bachelor's degree) and three other items referred to the applicant's national origin (Portuguese/British/Chinese). Finally, one filter item (e.g. “The candidate described in the résumé is applying to an entry-level accounting clerk job”) referred to common information presented in all conditions and was used to remove careless responses.

Data analyses

Before testing the hypotheses, descriptive statistics for the main variables within each experimental condition were computed. Chi-square and ANOVA tests were used to determine whether the sample characteristics were evenly dispersed. Experimental conditions did not differ from each other in participants' age (F(1,371) = 1.49, p = 0.13), gender (χ²(1,373) = 21.36, p = 0.49), participants' education (χ²(1,370) = 12.91, p = 0.29), previous residence abroad (χ²(1,367) = 13.58, p = 0.28), family living abroad (χ²(1,366) = 7.76, p = 0.74) and recruiting experience (χ²(1,341) = 16.08, p = 0.14), which support the assumption of randomisation. Additionally, the results for the manipulation checks ascertained that all participants correctly recalled the applicants' gender, education and national origin, as the means for the respective conditions were significantly higher than those for the opposite conditions.

Results

Descriptive analysis

Descriptive statistics for each experimental condition were computed before testing the hypotheses and are presented in Table 2. Next, descriptive statistics and Spearman correlations among the study variables appear in Table 3.

To test the hypothesis, respondents' demographics correlated with the criterion variables (i.e. gender, education, prior residence abroad, family living abroad and recruiting experience) were included as covariates.

Hypothesis testing

Multivariate covariance (MANCOVA) analyses were performed to detect mean differences across the 12 experimental conditions for each dependent variable. The data did not violate the assumption of homogeneity of variance, as Levene's test for equality of error variance was non-significant for the dependent variables at the p < 0.05 level, as well as Box's test of equality of covariance matrices at the p < 0.05 level.

Table 4 summarises the results, including the effect sizes reported by partial Eta square (ƞp²). The observed effects vary from small to moderate as per Richardson (2011) intervals, in that a partial Eta squared of 0.0099 is small, 0.0588 is moderate and 0.1379 is large. The observed power for the full models varied from 0.872 to 0.987, which is considered adequate.

Hypothesis 1 predicted a gender premium and higher ratings of (a) interpersonal skills for male candidates in comparison with female candidates (regardless of the level of education and national origin) and no gender differences in terms of (b) job skills and (c) job suitability. The results shown in Table 4 do not support a gender effect for interpersonal skills, which does not support H1a. However, women candidates outperformed men in job skills (M female = 3.50; M male = 3.34; F = 4,963, p = 0.027) and job suitability (M female = 3.69; M male = 3.52; F = 4,101, p = 0.044), which is contrary to hypothesis 1b and 1c. The findings reveal a gender premium favouring women's skills and job suitability to an entry-level accounting clerk job.

Hypothesis 2 predicted an education premium and higher ratings for the candidates with high education in all dependent measures compared to those with vocational education (regardless of gender and national origin). Following Table 4, the résumés of HE candidates scored higher in perceived similarity with the rater (M vocational = 1.99; M high-educt = 2.27; F = 7,090, p = 0.01) and job suitability (M vocational = 3.41; M high-educt = 3.81; F = 21,352, p = 0.000), but not on the other dependent variables which support hypothesis 2a and 2d, but not hypothesis 2b and 2c. The raters considered the candidates who have HE to be more similar to themselves than the less educated ones (even when raters did not belong to the higher education group), and while these candidates did not score higher on skills (i.e. interpersonal and job skills) they were considered more employable than the less educated. The findings reveal an education premium favouring HE candidates in job suitability.

Hypothesis 3 predicted a national origin premium and higher ratings in all dependent variables for the Portuguese candidates compared to the British and Chinese candidates (regardless of gender and education). Table 4 shows a marginally significant main effect of candidates' national origin only for perceived similarity. ANOVA results to determine the origin of this effect showed that the Chinese candidates scored lower than all the other candidates (M Chinese = 1.98; M British = 2.28; M Portuguese = 2.15; F = 2,833, p = 0.06), with no significant differences between the Portuguese and the British candidates. In other words, the Portuguese raters considered the British and the local candidates more similar to themselves than the Chinese candidates, but this difference was marginally significant. The results do not support H3, meaning no ingroup favouritism based on national origin could be confirmed.

Hypothesis 4 dealt with an expected ingroup derogation of the less-educated Portuguese candidates in (a) perceived similarity to the rater, (b) job skills and (c) job suitability, in comparison with the HE British candidates, whereas (d) no differences were expected for interpersonal skills. Table 4 shows a marginally significant interaction effect for education and national origin but only for candidates' interpersonal skills (F = 2.93, p = 0.06, np² = 0.06), which do not support H4. Surprisingly, high and less-educated Portuguese candidates scored alike in interpersonal skills (M _{NHE Portuguese} = 3.43; M _{HE Portuguese} = 3.46) and equally to HE British candidates (M _{HE British} = 3.45), as illustrated in Figure 1. Contrary to expectations, the HE Chinese candidates (M _{HE Chinese} = 3.69; SD = 0.77) and the non-HE British candidates (M _{NHE British} = 3.68; SD = 0.53) scored alike and higher than locals in interpersonal skills.

Finally, hypothesis 5 predicted outgroup favouritism and higher ratings of (a) perceived similarity; (b) job skills; and (c) job suitability for the HE British candidates in comparison to the equally qualified Chinese candidates. Given the results in Table 4 and the absence of significant interaction effects for these three dependent variables, this hypothesis is not supported. Other supplementary analyses are shown separately.

Discussion

The current study extends earlier research on multiple social categorisations in résumé screening by examining how the interplay of diverse social categories affects résumé screening and influences the inferences about the candidates' employability. Subjects rated one of twelve fictitious résumés designed to test the main and the interaction effects of gender, education and national origin. National origin included candidates from the local majority (Portuguese) and candidates of European ancestry (British), and a non-European minority (Chinese). All candidates were screened for the same entry-level accounting clerk job, for which a bachelor's degree is “good to have” but is not a basic requirement (Cooper and Robson, 2006).

The first unexpected finding was a generalised preference for female candidates, who were rated higher than men in job skills and job suitability, thus contradicting previous studies (Zschirnt and Ruedin, 2016; Pinto and Ramalheira, 2017; Pinto and Pereira, 2019). This gender premium was unrelated to candidates' interpersonal skills (contrary to predictions, female candidates did not score higher than men in these skills), so we cannot really conclude for a gender stereotype based on women's superiority in the skills that are “soft” (Andrews and Higson, 2008; Ellemers, 2018). However, the findings do not rule out the possibility of a gender stereotype. While the accounting profession has always been male-dominated (Del Baldo et al., 2019), and an accounting clerk job is gender neutral (Cooper and Robson, 2006), including locally (National Institute of Statistics, 2023a), we cannot dismiss an implicit gender bias that favours women without work experience, to perform administrative and clerical jobs (Cooper and Robson, 2006; Del Baldo et al., 2019). Conversely, such gender premium was not observed for the recruiters' subsample (see the supplementary findings), suggesting that these respondents were perhaps more thoughtful (and less judgmental) in assessing the candidates' credentials and fit to the job requirements.

The findings also reveal an education premium since the HE candidates were considered more similar to the raters (i.e. “ingroup-like”) and more job-suited but not necessarily more job skilled. The recruiters also recognised this higher job suitability. This education premium is not new (e.g. Green and Henseke, 2021; Pinto and Ramalheira, 2017) and confirms that highly qualified candidates are often favoured in the labour markets, even when jobs do not require a graduate qualification. An implication can be the progressive underemployment of graduates (Scurry and Blenkinsopp, 2011) and the brain waste of locals and skilled migrants (Farivar et al., 2019).

Unlike prior studies on hiring discrimination (Blommaert et al., 2014; Derous et al., 2012, 2016), including second-generation immigrants (e.g. Midtbøen, 2016), we found no ingroup favouritism towards nationals nor any explicit outgroup discrimination based on national origin. Given our design and the categorisations used, national and non-national candidates were equally employable among the HE cohort. In addition, HE Chinese female candidates modestly outperformed the other candidates (British and locals) in interpersonal skills but not in the other variables, for which no intersectional effects were observed. These results are consistent with previous findings (Alecu, 2019; Derous et al., 2009), specifically, the studies which use less distant minority groups (e.g. McGinnity and Lunn, 2011). Overall, the results revealed (1) a preference for female candidates to perform an accounting clerk job; (2) a preference for the HE, who looked more “ingroup-like” even among the less qualified raters; and (3) no significant differences pertaining the national origin of the candidates. Instead of an additive effect, we found no significant differences once the three social categories were combined.

The results confirm that intergroup bias in résumé screening can be reduced (Derous et al., 2009), especially when the résumés of minority candidates include other relevant cues, such as local academic credentials. In such circumstances, ingroup favouritism can be prevented. Finally, as these elements relate to person-job and person-culture fit (Jones et al., 2017), the raters might have inferred that the non-national candidates were as culturally and socially adjusted as locals, despite their different national origins, which can deter outgroup discrimination.

However, it remains unclear why ingroup favouritism towards local candidates was not observed, contradicting the assumptions of the social identity theory (Tajfel and Turner, 1986), as well as the predictions of SCM (Cuddy et al., 2009; Lewis and Sherman, 2003) and the theory of social similarity (Byrne, 1971; Grigoryan, 2020). Two explanations for this counterintuitive result can be offered. The first explanation is methodological. According to Greenwald and Pettigrew (2014), the methods used to test hiring discrimination and distinguish ingroup favouritism from outgroup discrimination have to (1) use a within-subjects design that permits comparison between ingroup and outgroup members; and (2) include measures that have “an unambiguous neutral point—a value that is neither favourable nor hostile” (Greenwald and Pettigrew, 2014, p. 680). Actually, none of these requirements were met. With a between-subjects design, we can confidently compare the preferences of groups of respondents but not each individual-level preference for one candidate versus another. Equally, the measures used to rate the favourableness towards the candidates, along with positive scales (and varied from 1 = very low competence/ability level to 5 = very high competence/ability level). Although this approach is similar to other studies that document ingroup favouritism, the results may actually document a “neutral” opinion of all candidates. The second explanation is theoretical and can be found in the approach of Gaertner and Dovidio (2000) to reduce intergroup bias, who emphasised the possible extension of ingroup boundaries by establishing “superordinate” or “dual identities”. Such an extension could have included as members of the ingroup all “candidates locally educated”. This reduction of intergroup bias through social re-categorisation (into a common identity or a dual identity) has been empirically documented (for a review, see Dovidio et al., 2009). It has been endorsed as a way to increase organisational diversity and inclusion (Gaertner and Dovidio, 2000). While acknowledging the many potential benefits of such an approach, we caution against its limitations.

On the one hand, such a “superordinate” identity can be viewed as a way to avert minority discrimination, including the discrimination of second-generation immigrants. On the other hand, it can be an obstacle to genuine organisational diversity because minorities are tolerated on the condition that they belong to a common “superordinated” (and dominant) category. Given the economic and social relevance of this issue, more research is required to capture the resumé’s characteristics that, considered in tandem, are more likely to signal such a superordinate identity, thus increasing the chances of employment; or, instead, are more likely to make the candidate more “outgroup like”, thereby increasing the chances of hiring discrimination.

Limitations and implications for future research

When interpreting the findings of this study, some limitations should be considered. This study used a between-subjects design, which cannot rule out contrasting attitudes towards individual ingroup–outgroup members (Greenwald and Pettigrew, 2014). Although survey experiments are common in this field, the respondents only rated their favourableness towards one candidate instead of comparing two or more. This can partially explain the absence of a clear ingroup favouritism-outgroup discrimination. Therefore, future mixed-methods studies can determine how the results stand for distinct minorities (e.g. (un)skilled Romany/black women/refugees/asylum seekers/digital nomads, etc.) and include other résumé’s cues, such as socio-economic status and attendance of public vs elite national and international schools.

Another limitation concerns the generalizability of the findings. While the sample was composed of experienced workers and was diverse in gender, education, recruitment experience and international exposure, participants were rather young on average (means varied between 29 and 35 years) and highly educated. Furthermore, some respondents had no recruiting experience, although supplementary analysis did not show differences in ratings according to this feature. From this perspective, future studies can determine how generalisable the findings are to other labour markets, minority groups, occupations and societies.

In addition, further research can examine how the perceived employability inferred from résumés’ content affects the selection process's subsequent steps. If, as shown, the job skills of locals and non-nationals do not differ once they are schooled locally, and thereby, they have equal chances of going ahead in the selection process, what are the chances of all being perceived as “ingroup-like” during a job interview? For instance, Wolgast et al. (2018) found that even the questions asked can vary depending on the interviewees' origin and anchoring and adjustment biases were found against stigmatised candidates (Buijsrogge et al., 2021). During a job interview, outgroup characteristics are more difficult to disguise, which might increase selection biases. On the one hand, recognising different skills and strengths may signal complementary competencies for the job, reinforcing the preference for diverse candidates (i.e. less ingroup favouritism and/or outgroup preference). On the other hand, such distinctiveness may lead to a “skill paradox” (Dietz et al., 2015), thus increasing the chances of hiring discrimination. These are some propositions requiring further empirical validation.

Finally, the results document a gender and education premium. Without engaging in outgroup-directed discrimination, respondents seem to have reproduced well-established social and gender norms, having privileged already favoured groups, such as the HE (Ford and Mellon, 2020) and the male candidates, depending on the type of job (Derous et al., 2015) and generally more adequate to higher-level accountant positions rather than office-clerk jobs (Del Baldo et al., 2019). It would be interesting to investigate these effects in other contexts and for gendered and other gender-neutral occupations, varying in other skill requirements.

In sum, this research makes several contributions to both theory and practice. Following earlier calls to contextualise the study of stereotypes in hiring (Czopp et al., 2015; Derous et al., 2021), we collected information about the general similarity of the candidates with the rater, in addition to their job skills and job suitability. The findings confirm that local workers can rate minority candidates, such as second-generation immigrants, as “ingroup-like” (Grigoryan, 2020, p. 1), challenging our conventional wisdom about minority discrimination.

Secondly, this study complements the use of the social identity theory as per the call of Fletcher and Beauregard (2022) and draws on the SCM (Cuddy et al., 2011) and the similarity-attraction theory (Byrne, 1971) to advance our understanding of how multiple social categories present in the résumé interplay and may decrease hiring discrimination in a specific context. The results document a gender and education premium, but these effects are not additive when these social categories combine with the national origin of the candidates. The results are consistent with other intergroup frameworks (e.g. Kang and Bodenhausen, 2015), including the common ingroup identity model (Gaertner and Dovidio, 2000; Dovidio et al., 2009), and provide opportunities for theory development. The findings confirm that multiple categorisations are context dependent and can be beneficial (Kang and Bodenhausen, 2015), thereby reducing bias in the hiring context. If, as shown, ingroup favouritism and outgroup discrimination can be reduced when other shared social categories come into focus, such as local education, then we can uncover more opportunities to decrease bias and increase the individuated analysis in résumé’s screening to promote organisational diversity.

Finally, this study offers a methodological addition to previous research on résumé screening by demonstrating the role played by the interaction between the rater and the candidate's characteristics. Previous research employing field experiments could hardly control for recruiter-based effects. Furthermore, the use of field experiments, along with the predominant use of low-status minority groups, might explain the over-discrimination of minorities traditionally reported. Instead, this study shows that job candidates hardly belong to one singular social category (i.e. male/female, educated/non-educated, local/foreigner, etc.), so more nuanced stereotypes can occur. This finding is striking to labour markets where locals are particularly aware of other national groups through emigration or immigration flows. In such a context, the findings show that minority hiring discrimination can be reduced.

Managerial relevance

This research also holds practical implications for candidates, recruiters, employers and European policymakers. A practical implication for female applicants is the validation that a premium towards them can voice implicit discrimination if, as shown, female candidates are more skilled and suited to an entry-level accounting clerk job than male candidates with equal credentials. Another practical implication to candidates and employers is the confirmation of a HE premium, even for jobs that someone with vocational training can perform. This risks leading to underemployment and brain waste among recent graduates. Another practical implication for second-generation immigrants, including those from more distant national origins, is that résumé discrimination can be prevented once they get local education and present themselves as highly qualified for the job. This will likely make them look “ingroup-like”, raising their employment chances.

The findings suggest that decision-makers must endorse a truly unbiased assessment of the applicant's job fit to promote organisational diversity instead of endorsing group stereotypes. Furthermore, the findings suggest that even the most positive expectations about the candidates can be discriminatory since they can condemn women and less-educated people to a stereotyped view instead of an individualised job-fit assessment. To recognise and appreciate multiple social categories in others, recruiters and managers might benefit from training so they can learn to recognise and appreciate their own diversity to embrace the diversity of others.

Employers are also urged to focus on factual résumé content that is relevant to the job requirements and limit the unrelated résumé-based inferences that can exclude adequate candidates and preclude workforce diversity. Finally, the policymakers aiming to contribute to an agenda for decent work and economic growth, including the reduction of inequalities, are worth learning that (1) gender bias can persist in the presence of comparable job skills, and even a gender premium can constrain women and men to jobs that are gender-related; (2) ingroup favouritism and hiring discrimination based on national origin can be reduced, at least for the HE second-generation immigrants that are educated locally. Hence, training hiring managers and recruiters to recognise and avoid bias and invest in the education of underrepresented groups can help increase diversity.

Conclusion

Given the importance of understanding how multiple social categorisations can influence hiring discrimination, this research investigated the interplay of gender, education and national origin in the résumé screening of young candidates applying for an entry-level accounting clerk job. The study found a gender premium, in that women scored higher in job skills and job suitability, and an education premium, as higher-educated candidates outperformed the less skilled even for an occupation that does not require higher education. No ingroup favouritism nor outgroup discrimination based on national origin was observed, and no significant interaction effects were found. These findings suggest that the preferences for women and HE candidates of all origins are largely independent. While these results can be specific to the local setting and job scenario, our approach highlights the value of studying stereotypes in context. It points out that discrimination in résumé screening depends on cues other than national origin. In the European context, and for second-generation British and Chinese immigrants, bias can be avoided once the résumé reports local academic credentials. A key conclusion of this study, however, is that outgroup candidates can be valued for their education and job skills. Still, this distinctiveness is insufficient to judge them as more employable than ingroup candidates.

Figures

Figure 1

Interpersonal skills of local and non-local candidates, according to gender and education

Table 1

Sample demographics per experimental condition

Experimental condition	N	Age		Gender		Education		Prior residence abroad	Family living abroad	Recruitment experience
		M	SD	Female	Male	< HE	HE
Condition A (English × Male × HE)	32	34.10	11.20	65.60%	34.40%	16.10%	83.90%	26.70%	60.00%	55.60%
Condition B (English × Female × HE)	31	35.40	9.40	41.90%	58.10%	12.90%	87.10%	25.80%	58.10%	67.70%
Condition C (English × Male × NHE)	31	31.10	7.60	54.80%	45.20%	12.90%	87.10%	41.90%	48.40%	39.30%
Condition D (English × Female × NHE)	31	29.10	6.30	67.70%	32.30%	6.70%	93.30%	29.00%	58.10%	54.80%
Condition E (Portuguese × Male × HE)	32	32.90	9.80	65.60%	34.40%	9.40%	90.60%	18.80%	50.00%	63.30%
Condition F (Portuguese × Female × HE)	32	31.10	7.60	46.90%	50.00%	18.80%	81.30%	40.60%	56.30%	46.90%
Condition G (Portuguese × Male × NHE)	30	30.70	7.40	60.00%	36.70%	20.00%	80.00%	36.70%	70.00%	50.00%
Condition H (Portuguese × Female × NHE)	31	30.90	9.10	51.60%	45.20%	10.00%	90.00%	48.30%	69.00%	48.10%
Condition I (Chinese × Male × HE)	32	32.00	8.90	50.00%	43.80%	6.30%	93.80%	43.80%	56.30%	34.40%
Condition J (Chinese × Female × HE)	30	29.40	6.60	53.30%	46.70%	13.30%	86.70%	24.10%	60.70%	37.50%
Condition K (Chinese × Male × NHE)	30	30.30	6.40	66.70%	33.30%	0.00%	100.00%	27.60%	48.30%	39.10%
Condition L (Chinese × Female × NHE)	31	31.90	8.40	51.60%	48.40%	3.20%	96.80%	41.90%	67.70%	64.30%
Overall Sample	373	31.60	8.40	56.30%	42.40%	10.80%	89.20%	33.80%	58.50%	50.40%

Note(s): HE–Higher Education, NHE–No Higher Education, M average, SD standard deviation

Source(s): Authors own creation

Table 2

Descriptive statistics per experimental condition

Experimental condition	N	Perceived similarity		Interpersonal skills		Job skills		Job suitability
		M	SD	M	SD	M	SD	M	SD
Condition A (English × Male × HE)	32	2.38	1.01	3.40	0.78	3.41	0.76	3.82	0.72
Condition B (English × Female × HE)	31	2.48	1.03	3.50	0.44	3.57	0.57	3.80	0.68
Condition C (English × Male × NHE)	31	2.23	1.15	3.67	0.57	3.46	0.71	3.41	0.96
Condition D (English × Female × NHE)	31	2.03	0.84	3.69	0.49	3.54	0.52	3.59	0.72
Condition E (Portuguese × Male × HE)	32	2.25	1.05	3.59	0.52	3.44	0.70	3.74	0.79
Condition F (Portuguese × Female × HE)	32	2.34	1.04	3.33	0.54	3.22	0.65	3.83	0.66
Condition G (Portuguese × Male × NHE)	30	1.93	0.87	3.38	0.70	3.21	0.69	3.48	0.96
Condition H (Portuguese × Female × NHE)	31	2.06	0.96	3.48	0.59	3.38	0.66	3.35	0.90
Condition I (Chinese × Male × HE)	32	1.84	0.85	3.48	0.83	3.38	0.73	3.65	0.71
Condition J (Chinese × Female × HE)	30	2.33	1.15	3.93	0.62	3.73	0.63	3.96	0.67
Condition K (Chinese × Male × NHE)	30	1.77	0.68	3.40	0.50	3.25	0.57	3.13	0.83
Condition L (Chinese × Female × NHE)	31	2.00	1.00	3.48	0.64	3.40	0.53	3.53	0.69
Overall Sample	373	2.14	0.99	3.53	0.63	3.41	0.66	3.61	0.80

Note(s): HE–Higher Education, NHE–No Higher Education, M average, SD standard deviation

Source(s): Authors own creation

Table 3

Descriptives and Spearman correlations for the main variables

Variable	M	SD	1	2	3	4	5	6	7	8	9	10	11	12	13
Respondents' Demographics
1. Age	31.60	8.42		0.17	−0.10	0.06	0.05	0.18**	0.02	0.09	0.01	−0.06	−0.07	−0.06	−0.06
2. Gender	−	−			−0.04	0.11*	0.09	0.08	−0.08	0.05	0.00	−0.03	-0.12*	−0.10	−0.10
3. Education	−	−				0.14**	−0.01	0.05	0.00	−0.06	−0.12*	−0.10*	−0.05	−0.07	−0.04
4. Prior residence abroad	−	−					0.15**	0.02	−0.03	−0.08	0.01	−0.07	−0.09	-0.12*	−0.08
5. Family living abroad	−	−						0.09	−0.06	−0.04	0.02	−0.04	−0.10	-0.12*	−0.08
6. Recruitment experience	−	−							−0.07	0.01	0.06	−0.07	−0.09	-0.15**	−0.08
Independent variables
7. Candidate's gender	−	−								0.01	−0.01	−0.07	−0.05	−0.09	−0.07
8. Candidate's education	−	−									0.01	0.13*	0.02	0.08	0.23**
9. Candidate's national origin	−	−										0.07	−0.08	−0.07	0.02
Dependent variables
10. Perceived similarity	2.14	0.99											0.25**	0.33**	0.29**
11. Interpersonal skills	3.53	0.63											(0.884)	0.65**	0.32**
12. Job skills	3.41	0.67												(0.869)	0.46**
13. Job suitability	3.61	0.80													(0.897)

Note(s): Spearman coefficients. Two-tailed. Significant at: *p < 0.05. **p < 0.01. n = 305–374. Cronbach's alpha estimates in parentheses, along the main diagonal. Gender (0 = Female; 1 = Male); Education (0 = No High Education, 1 = High Education); Prior residence abroad (0 = No, 1 = Yes); Family abroad (0 = No, 1 = Yes); Recruitment experience (0 = No, 1 = Yes); candidate's gender (0 = Female; 1 = Male), candidate's education (0 = No High Education, 1 = High Education); candidate's national origin (1 = Chinese, 2 = English, 3 = Portuguese)

Source(s): Authors own creation

Table 4

MANCOVA results for the dependent variables after incorporating the respondents' demographics as covariates

Source/Dependent variable	Perceived similarity				Interpersonal skills				Job skills				Job suitability
	df	MS	F	ƞp2	df	MS	F	ƞp2	df	MS	F	ƞp2	df	MS	F	ƞp2
Gender	1	0.31	0.33	0.00	1	1.03	2.78	0.01	1	1.10	2.57	0.01	1	1.10	1.79	0.01
Education	1	1.27	1.34	0.00	1	0.67	1.80	0.01	1	0.77	1.79	0.01	1	0.00	0.01	0.00
Prior residence abroad	1	0.03	0.03	0.00	1	0.05	0.14	0.00	1	0.38	0.89	0.00	1	0.09	0.14	0.00
Family living abroad	1	0.34	0.35	0.00	1	1.11	3.01+	0.01	1	2.37	5.52*	0.02	1	1.18	1.91	0.01
Recruitment experience	1	2.23	2.35	0.01	1	0.38	1.02	0.00	1	2.43	5.66*	0.02	1	2.39	3.87*	0.01
Candidate gender (A)	1	3.12	3.27+	0.01	1	0.87	2.37	0.01	1	2.13	4.96*	0.02	1	2.53	4.10*	0.01
Candidate education (B)	1	6.75	7.09***	0.02	1	0.04	0.10	0.00	1	0.72	1.67	0.01	1	13.17	21.35*	0.06
Candidate nationality (C)	2	2.49	2.61+	0.02	2	0.66	1.78	0.01	2	0.85	1.98	0.01	2	0.19	0.31	0.00
A × B	1	0.10	0.10	0.00	1	0.01	0.03	0.00	1	0.02	0.05	0.00	1	0.09	0.15	0.00
A × C	2	1.09	1.15	0.01	2	0.77	2.09	0.01	2	0.43	1.01	0.01	2	1.28	2.08	0.01
B × C	2	0.34	0.36	0.00	2	1.08	2.93+	0.02	2	0.51	1.19	0.01	2	0.19	0.31	0.00
A × B × C	2	0.01	0.01	0.00	2	0.85	2.31+	0.01	2	0.42	0.97	0.01	2	0.69	1.11	0.01

Note(s): MANCOVA = multiple analysis of covariance. df degrees of freedom, MS Mean square, F F ratio, ƞp2 partial eta squared effect size. All values were computed for the corrected model. Significant at: +p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001

Source(s): Authors own creation

Table A5

Summary of stepwise regression analysis

Effects between conditions	Perceived similarity of the candidate
	Step 1	Step 2
Step 1 - Respondents Demographics
Gender
Education	0.11*
Prior residence abroad
Family living abroad
Recruitment experience
Effects of Education (Hypothesis 3)
Step 2 - Candidate Attributes
Gender (A)		0.23**
Education (B)
National origin (C)
A × C	0.57	0.23
B × C
Overall F	1.300	1.663*
R2	0.02	0.06
Adjusted R2	0.01	0.02
Change in R2	0.02	0.04

Note(s): Significant at: +p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001; standardized β coefficients are reported for each step and after Z-score transformation. Gender (0 = Female; 1 = Male); Education (0 = No High Education, 1 = High Education); Prior residence abroad (0 = No, 1 = Yes); Family abroad (0 = No, 1 = Yes); recruitment experience (0 = No, 1 = Yes), National origin (1 = Chinese, 2 = English, 3 = Portuguese)

Source(s): Authors own creation

Appendix

Supplementary Findings

Interaction effects

Table 4 also indicates a marginally significant interaction effect for gender × education × national origin pertaining to interpersonal skills. ANOVA differences between the conditions and LSD post-hoc tests showed that the HE Chinese female scored higher in interpersonal skills than all other candidates, except the non-HE British candidates (regardless of gender). These results do not support H4. However, this finding suggests the possibility of outgroup favouritism in the case of positive counter-stereotypical information. When the résumé information (i.e. education and national origin) contradicts a minority's stereotype (i.e. the Chinese are locally seen as unskilled, underemployed and cold), one can observe outgroup favouritism even for a dimension relevant to the ingroup.

Table 4 indicates that some respondents' demographics influence the assessment of the candidates' job skills and job suitability. Subsequent ANOVA showed that raters with family living abroad were more severe in assessing the job skills of the candidates (M _{No family abroad} = 3.53; M _{family abroad} = 3.34; F = 6,659, p = 0.01). Conversely, the respondents with recruiting experience were also more severe in rating the job skills of the candidates (M _{No recruit} = 3.52; M _recruit = 3.32; F = 7,375, p = 0.007), and their job suitability (M _{No recruit} = 3.70; M _recruit = 3.54; F = 3,502, p = 0.05).

Respondents demographics

Although it was not hypothesised, the analyses were repeated for the subsample of recruiters (i.e. the 172 respondents with recruiting experience). The number of responses per condition varied from 9 to 21, so the observed power decreased and varied from 0.501 to 0.941. All in all, the results were similar to the ones previously reported, except that there was no gender effect. Specifically, there was an education effect only for job suitability (M _vocational = 3.38; M _high-educt = 3.64; F = 3,823, p = 0.05), and there was a marginally significant effect of gender × education × national origin for candidates' interpersonal skills (df = 2, MS = 0.98, F = 2,571, p = 0.08, np² = 0.03), in that the female HE Chinese scored the highest in interpersonal skills. These results confirm no ingroup favouritism towards local candidates nor outgroup discrimination of local minorities (regardless the gender and education).

Perceived similarity of the candidate

Although it was not hypothesised, the predictors of the perceived similarity of the candidate were explored to understand which résumé content prompted this perception. Given that one can perceive target individuals who represent multiple social categories as “ingroup like” or “outgroup like” (Grigoryan, 2020, p. 1), despite their specific attributes, one examined which candidates were more “ingroup like”.

Table A5 presents the results of a stepwise regression analysis, including as inputs the raters' characteristics (to determine if certain raters were more prone to make similarity attributions) and then the résumés attributes.

As shown in Table 5, the less educated raters were more prone to find the candidates attractive and similar to themselves, which was also influenced by the candidates’ education. The findings confirm that Portuguese respondents considered the HE candidates more identical to themselves, despite the candidates’ gender and national origin.