Abstract
The role of employer discrimination in labour market matching is often acknowledged but challenging to quantify. What part of the ‘ethnic penalty’ in the labour market is due to recruitment discrimination? This experiment, the first of its kind in Ireland, explicitly measured this by sending out nearly 500 equivalent CVs from Irish and minority candidates in response to advertised vacancies in the greater Dublin area. We find that candidates with Irish names are over twice as likely to be called to interview as are candidates with an African, Asian or German name. This discrimination rate is high by international standards, and does not vary between minority groups. We develop the discussion of the role of prejudice and stereotypes in discrimination in this article, arguing that our findings may be linked to the fact that Ireland is a ‘new immigration’ country, with no established minority groups and a cohesive national identity.
Introduction
A key concern of labour market sociology in the past decades has been how individuals get assigned to particular jobs, and the impact of this on labour market stratification more generally. Minority or migrant status is a key stratifying mechanism. Recent research has highlighted persisting inequality between migrant or minority groups and the native population across a range of labour market outcomes. Unemployment is typically higher among immigrants than among natives (Fleischmann and Dronkers, 2010). Minority groups may be disadvantaged in terms of working conditions (Turner, 2010) and precarity (Anderson, 2010), in Britain and other European countries (Heath and Cheung, 2007) and in the USA, though important variations in the performance of different groups do exist. In Ireland, where immigration is a recent phenomenon, there are indications that a similar pattern of migrant disadvantage may have developed (Barrett and McCarthy, 2007; Turner, 2010). Part of this labour market disadvantage could be because of group differences in abilities, skills and experience, though previous research suggests that once these are controlled for, a residual difference still exists. To what extent is this employer discrimination?
The typical approach to measuring discrimination is to compare the wages or jobs of majority and minority populations, statistically controlling for differences in education and experience (human capital characteristics). The remaining difference between groups is often attributed to discrimination, though this is problematic given that some relevant human capital differences may be incompletely measured. Using surveys to ask people directly about their experience of discrimination is another method, though here reports of discrimination can vary depending on the perspective of the respondent, their expectations and the information available to them (Russell et al., 2008).
Given serious flaws in both these methods, a more unusual alternative is to measure discrimination in recruitment directly using an experiment. We designed such an experiment to test the recruitment process by creating fictional individuals with identical human capital, but with names that invoke different ethnic backgrounds. If the response to each CV is the same, this suggests no discrimination; if there is a difference, it suggests otherwise. Discrimination is understood as the unequal treatment of an individual on the basis of group membership. The experiment thus focuses on individual recruitment decisions, applying experimental principles (matching and random assignment) to actual job applications. While there are many junctures at which employers could generate inequality – recruitment, pay, promotion – this article focuses on the recruitment process as the first step. If people do not get the jobs, they will never get paid or promoted. What distinguishes this experiment as scientific is the number of responses collected: almost 500 CVs were sent out in order to be sure that the results were systematic and not due to chance. It allows us to explicitly measure the extent of recruitment discrimination for the first time in Ireland.
Why Ireland? What makes Ireland interesting is that a large and diverse group of immigrants rapidly entered a relatively small labour market that was previously almost exclusively white and Irish. A comparison of the 1996 and 2006 censuses shows that 1.7 per cent of the resident population in 1996 was born outside Ireland and the UK, but by 2006 this figure was 8.2 per cent. Thus we have a society with no established foreign-born ethnic minority groups, employers with limited experience of hiring ethnic minorities and a very cohesive national identity. This puts Ireland in a similar position to other countries in the periphery of Europe that have experienced recent rapid immigration (Bail, 2008). In the next section we explore some of the implications of the recent nature of immigration, and how it might influence the recruitment of minorities.
This article contributes to the literature in a number of ways: empirically by providing clear and convincing evidence of the extent of discrimination in recruitment in Ireland, a ‘new immigration’ country; methodologically by highlighting the advantages of using experiments to explore the issue; theoretically by suggesting that a small amount of bias could result in substantial discrimination and inequality in allocation to jobs. In the remainder of the article we consider theoretical perspectives on discrimination and how they might apply to the Irish case. We then discuss the measurement of discrimination and the experimental design. The results of the experiment are presented, followed by a discussion which reflects on the findings and their implications for both theoretical approaches and future experiments.
Conceptualizing discrimination
Why then would employers discriminate on the basis of ethnic background? One influential perspective is that discrimination is based on prejudice. In Becker’s (1971) ‘taste for discrimination’ theory, he argues that some individuals, be they employers, co-workers or customers, have a ‘taste for discrimination’, while some are indifferent to race. Prejudiced employers will prefer majority candidates, and impose a ‘discriminatory psychic penalty’ for minority applicants (OECD, 2008). This theoretical approach also implies that in jobs with face to face contact, like sales, discrimination will be higher, as employers take into account customer prejudice (Darity and Mason, 1998). Clearly, the greater the extent of prejudice against minorities, the greater the discrimination in recruitment will be.
That said, in the case of equivalent CVs, a very small amount of prejudice held by many employers could result in a high rate of discrimination (see Cohn, 2000). Typically, employers recruiting through job advertisements will select only a minority of the candidates who respond, perhaps a small minority. Given an initial (subjective) assessment of suitability for each candidate, they hence must apply a threshold that is towards the tail of the distribution of suitability across candidates. If so, where even a small amount of taste discrimination enters the assessment of suitability, it could translate into a substantial penalty in terms of the probability of being selected.
The idea that prejudice is at the root of discrimination is not new: how prejudice develops and manifests itself has long been the concern of sociologists and social psychologists. One example of prejudice, drawing on the social identity perspective, is the notion of in-group favouritism. This implies the extension of trust, positive regard and empathy to the in-group but not out-group members (Hewstone et al., 2002). Results across hundreds of studies show how individuals evaluate in-group members more positively, give preference to the in-group in the allocation of resources, and seek to maximize the difference in allocation between the in-group and the out-group (Al Ramiah et al., 2010). Given that group membership in this model is not a deeply held attachment, does not contain an active element of aggression and operates in the context of equality norms, these findings of discrimination are salient. They suggest that relatively mild in-group bias can result in substantial discrimination in the allocation of resources, in this case jobs. Again a relatively small amount of prejudice among many individuals could give rise to high levels of discrimination.
While overall attitudes to migrants and migration have been relatively positive in comparative context, in-group bias has resonance in the Irish context given the almost negligible experience of racial/ethnic minorities, combined with a strong, cohesive national identity. Bail (2008) argues that religious/racial boundaries are relatively strong in Ireland. In this regard Ireland is similar to other peripheral European countries such as Spain, Portugal, Italy, Greece and Hungary, where immigration is also relatively recent. Bail (2008) links these strong boundaries to the relative homogeneity of these new immigration countries. Regarding national identity, there is evidence that compared to those living in other European countries, people living in Ireland place much greater emphasis on the importance of the Catholic religion and being born in Ireland in their understanding of Irish nationality (Davis, 2003, using ISSP data). They also exhibit very high levels of national pride (Davis, 2003, using Eurobarometer data). Following this line of argument, it could be that in-group bias is relatively common in Ireland, and that discrimination does not vary between minority groups.
Yet it is important to stress that not all who are prejudiced go on to discriminate. Other factors are associated with discrimination, as actors operate in social settings. Strong norms of equality, profit motivation, diversity training, and other organizational features may mean that employers do not discriminate. Sociologists have considered the role of organizational structures in mediating the prejudice and stereotypes of actors (Reskin, 2000), and also what are loosely termed ‘structural factors’– country-level policies and practices that contribute to the systematic disadvantage of certain groups (Pager and Shepherd, 2008).
Alternative models of labour market discrimination challenge the notion that prejudice is at the root of discrimination. For statistical discrimination models, imperfect information about workers’ abilities constitutes the key rationale for discrimination (e.g. Arrow, 1973). Hiring decisions make use of prior beliefs, true or false, about the distributions of productivity associated with different groups. Consequently, statistical inferences made by employers might lead to assessments of productivity that are partly based on group membership, such that candidates who appear individually equivalent are assigned different expected productivity.
Employers’ inferences may vary in sophistication, perhaps involving fairly basic heuristics. Group-based generalizations are used as a heuristic to provide guidance about the expected productivity of individuals from a given group. 1 Thus hiring decisions are in part based on prior beliefs about a group, be they true or false. The effects of these heuristics may vary depending on the availability of and attention to person-specific information (such as that provided on the CV), which may act with or may not override initial expectations. They may also depend on the nature and level of qualifications: more specific and higher qualifications might reduce uncertainty and the need to use group membership as a proxy for ability, therefore reducing discrimination. At its most extreme form, employers might use the name as a heuristic for determining whether to make the effort to call the candidate to interview. They do not even read past the name to process other characteristics relating to the candidate’s qualifications and experience that might override initial expectations and reveal that the candidate’s qualifications and experience are equivalent (Bertrand and Mullainathan, 2004).
Stereotypes are typically defined as beliefs about a group and can be over-generalized, inaccurate and resistant to change. They are distinguished from prejudice as cognitive, not affective. Research has examined how stereotypes can be modified by information about the individual, as well as the conditions under which they are resistant to change (Fiske, 1998). In cases of decision-making under uncertainty, racial bias may play a role. Stereotypes may cause employers to filter information that preserves expectation, evaluating qualifications differently depending on the group membership. The role is much stronger if employers do not read past the name: there is no chance to modify the stereotype with individualizing information. Racial bias would certainly lead us to expect differences in the discrimination rate between white and non-white minority groups in this experiment.
Previous experiments investigating racial discrimination have been conducted in a number of countries, with a range of minority groups, and have found remarkably consistent evidence of discrimination against minorities (for reviews, see OECD, 2008; Riach and Rich, 2002). A second consistent finding is that of lower discrimination against white immigrants in predominantly white societies, in the studies which tested this (e.g. Booth et al., 2010). These findings suggest that while immigrants per se face discrimination in recruitment, the extent of discrimination faced depends on the colour of their skin. The findings on variations in discrimination across the labour market are far less consistent, in terms of skill levels, firm size or neighbourhood effects (Bertrand and Mullainathan, 2004; Carlsson and Rooth, 2007).
This discussion of the previous literature has suggested a number of plausible outcomes of the experiment, given the recent nature of migration in Ireland, and we develop a number of hypotheses in the following section that draw on this discussion.
Research hypotheses
A key question to emerge from the consideration of prejudice and stereotypes is whether discrimination varies among minority groups. We test for the presence of discrimination in recruitment against Asians, Africans and Germans. 2 As is typical in experiments of this nature, we use fictitious applicants with distinctive Asian, African and German names and compare them to applicants with typical Irish names. 3 As to what the differences between groups might be, or the overall rate of discrimination, the analyses in this article are guided by three research questions and concomitant hypotheses:
Firstly, are there any differences in responses to the minority candidates and the Irish candidate? Given both the theoretical discussion, previous international findings from field experiments and differential outcomes in the Irish labour market, our first hypothesis (1a) is that we will find discrimination in recruitment against minority candidates. While attitudes are not particularly negative towards migrants and migration in Ireland (Hughes et al., 2007), the discussion above suggests that even relatively mild prejudice or bias could result in high rates of discrimination, if held by a large number of recruiters (1b).
Secondly, is there any variation in the extent of discrimination between the minority groups? If racial bias – either prejudice or stereotypes – directed towards non-white groups is playing a strong role in recruitment decisions, hypothesis (2a) is that we will find higher discrimination in recruitment against non-white immigrants (African, Asian) than white immigrants (European). This would also be consistent with the findings from other countries. Alternatively, if it is a case of ‘in-group favouritism’, hypothesis (2b) is that the discrimination rate would be rather similar between all three groups.
Thirdly, does discrimination vary across the labour market in Ireland? We examine three occupations - administration, accountancy and sales. From the ‘taste for discrimination’ perspective we might expect that discrimination would be higher in sales, an occupation with frequent customer contact, as employers account for customer prejudice (hypothesis 3a) (Darity and Mason, 1998). From the statistical discrimination approach we might imagine that discrimination would be lower in accountancy jobs, where qualifications are somewhat more specific and credentials play a greater role (hypothesis 3b). If neither of these effects is operating, the discrimination rate might be rather similar across occupations (hypothesis 3c).
Measuring discrimination: the experimental design
Field experiments in recruitment have been used to investigate discrimination on the basis of gender (Fasang, 2006); family status (Corell et al., 2007); age (Bendick et al., 1999) and recently social class (Jackson, 2009), as well as race. These experiments retain key elements of the laboratory experiment (matching, random assignment), but apply them to real contexts (like job searches) to measure outcomes (Pager and Shepherd, 2008).
Personal approaches, i.e. the use of matched pairs of testers who pose as job applicants in real job searches, have been criticized for not being able to demonstrate the equivalence of testers (Heckman and Siegelman, 1993). The present experiment circumvents this criticism by using a written approach: correspondence testing. In the experiment, two fictional individuals, who are identical on all relevant characteristics other than the potential basis of discrimination, apply for the same jobs. Responses are carefully recorded, and discrimination or the lack thereof is then measured as the extent to which one applicant is invited to interview relative to the other applicant. Correspondence testing does have limitations. Probably the most salient weakness is that using this method, only jobs requiring a written application are available for testing. Correspondence tests are also confined to the first stage of the hiring process, i.e. selection to interview. However, the latter need not be a serious problem, as evidence from studies conducted by the ILO suggests that most discrimination occurs at the initial stage (i.e. selection for interview), not at the stage ‘interview to job offer’ (Bovenkerk, 1992).
Designing a field experiment proved challenging in Ireland, where, given the small labour market, the risk of detection is greater and the number of job vacancies lower. 4 A number of general principles informed many operational decisions. Firstly, we wished to avoid detection. Thus we only ever sent out two CVs per vacancy, not up to six, as in other countries, to avoid arousing suspicion. CVs were designed to be equivalent but not identical. Secondly, an experiment of this nature raises ethical questions, primarily due to the degree of deception involved, as employers are unaware that they are part of the experiment. This project went through a rigorous ethics procedure before the project commenced. The primary defence of the deception involved is that direct evidence of labour market discrimination is not available by any other technique (Riach and Rich, 2004). Steps were also taken in the design to minimize the inconvenience, costs and damage to the reputation of employers. Thirdly, due to time and financial constraints, occupations requiring written applications that had many vacancies were targeted to reduce the time required to conduct the experiment: lower administration and lower accountancy and retail sales positions. 5 Finally, we wanted to create high-quality CVs that were realistic and plausible for the jobs advertised. The higher the rate of response to CVs, the fewer CVs need to be sent out.
For each occupation, two equivalent CVs were developed that included personal and contact details, education, work experience, hobbies/interests and other skills. The receipt of an Irish Leaving Certificate also indicated the English-language proficiency of the minority candidates. In order to avoid detection CVs were not identical, but within each job category personal and employment characteristics were matched between the two CVs. Essentially, applicants differed only in their ethnically distinctive names. The effectiveness of this field experiment depended on employers recognizing the ethnicity of job applicants, so a small pre-test was conducted to identify names which were most readily identifiable as Irish, African, Asian or German. 6
The first stage in the application procedure was to identify vacancies in each of the three target areas: lower administration, lower accountancy and retail sales. Jobs advertised by recruitment agencies were excluded as the risk of detection was considered too high, as were positions that required detailed application forms. Each company was only included once in the study. Within these constraints, all advertised vacancies were applied for in the greater Dublin area.
Once a vacancy had been identified, two matched CVs were sent by email to the advertising employer. One CV was always from the Irish candidate and the other one was alternated between the African, Asian or German candidate. For equivalence, the CVs were rotated across identities, so the minority candidates and Irish candidates each got assigned CV1 and CV2 in turn. There were no significant differences in type of response to CV1 and CV2 in any of the three occupations, indicating that the CVs were indeed equivalent. 7 The first applications were sent out in early March 2008 and the final applications at the end of September 2008. All response information was recorded confidentially, and once an employer responded, the research team promptly declined any invitations to interview and terminated the recruitment process. Of the 240 job advertisements responded to, responses were received in respect of 111 in total (including rejections). Response rates were similar across occupations and minority groups. Approximately 39 per cent of jobs applied for received at least one favourable response; this relatively high rate of positive responses is indicative of the quality of the CVs. 8
Results: discrimination in recruitment
What were the responses to the matched pairs of fictional applicants? Table 1 presents a breakdown of responses to the 240 pairs of matched job applications. Of these, no response was received or both candidates were rejected in 147 cases. The remaining 93 cases are classified into three categories, distinguishing those where both candidates were invited to interview, those where the candidate with the ‘Irish’ name was asked to interview and the candidate with the ‘minority’ name was not, and those where the minority candidate was invited to interview but the Irish candidate was not.
Classification of responses to matched job applications
Do we find evidence of discrimination against minority candidates?
The first research question was whether there are any differences in responses to the minority candidates and the Irish candidate. Table 1 shows that the incidence of an interview being granted to the Irish candidate but not the minority candidate is substantially higher than the incidence of an interview being granted to the minority candidate but not the Irish candidate. If we consider these cases of non-equal treatment as discrimination, discrimination against the non-Irish candidate occurred in 55 cases, while discrimination against the Irish candidate occurred in just 15 cases.
Some discrimination against majority white applicants is typically found in experiments of this nature, and is usually ascribed to a randomness/inefficiency in the recruitment process (Riach and Rich, 2002). This is why all estimates of discrimination we discuss below are of net discrimination (see row 5), i.e. discrimination against the minority minus discrimination against the Irish candidate. In this case net discrimination for the total sample is 40.
A standard measure of the extent of discrimination is the ‘net discrimination rate’, which measures net discrimination as a proportion of those instances where at least one candidate was invited to interview (row 6). This discrimination rate, while commonly used, is not unproblematic as a measure of discrimination. Should we measure discrimination as a percentage of all applications sent or of applications for which at least one candidate was invited to interview? In Tables 1 and 2, we have conformed to practice elsewhere and used the latter, but there is controversy about this in the field experiment literature (Heckman and Siegelman, 1993; Riach and Rich, 2002).
Our preference is to highlight the relative risk of being asked to interview, or the ‘relative callback rate’, which is presented in the final row of Table 1. This is defined as the probability that the Irish candidate is asked to interview relative to the probability that the minority candidate is invited. More simply, it tells us how much more likely it is that the Irish candidate is asked to interview. In the present case, from the final column of Table 1 we can see that Irish candidates are invited to interview a total of 78 times, while minority candidates are invited a total of 38 times. For a given denominator N, the relative callback rate is:
That is, in our experiment candidates with an Irish name are over twice as likely to be asked to attend an interview as are candidates with an African, Asian or German name. This is the scale of discrimination encountered.
Is this discrimination statistically significant and does it differ between groups?
Although, at first sight, this disparity is striking, are the differences in treatment between Irish and minority candidates statistically significant?
For a given level of positive responses to candidates’ applications, the appropriate comparison is between those cases where the Irish candidate is invited and the minority candidate is not and those where the minority candidate is invited and the Irish candidate is not. The null hypothesis is that each of these cases occurs with equal probability; that is, that there is no greater likelihood of observing discrimination against minority candidates than against Irish ones.
If the total number of these cases of discrimination is m, and the number of cases where the minority candidate is asked to interview and the Irish one is not is δ, then we can use the binomial distribution B(m, p) to calculate the probability of observing no more than δ cases, given the null hypothesis that discrimination is as likely to operate in both directions (p = ½):
The result of this analysis is a p-value that equates to the probability that the data could have been observed if, in reality, there were no greater likelihood of discrimination against the minority candidate than of discrimination against the Irish candidate. The analysis can be done with the three different minorities pooled into a single group, or separately for each minority. The resulting p-values are given in Table 2, which also presents responses to the CVs for each group separately. From the figures in Table 2, we can conclude that the higher incidence of discrimination affecting minority candidates is strongly statistically significant. Indeed, the tiny probability in the final column reveals that the chance of observing these responses, if there were in fact no discrimination in the real world, is less than one in a million. So our first hypothesis (1a) is confirmed: there are clear differences between Irish and minority candidates in their chances of being called to interview.
Responses to matched job applications by minority group with significance tests
Table 2 also shows that discrimination against each of the three minorities considered separately is greater than discrimination against the Irish candidates. Furthermore, according to conventional criteria for statistical significance, the level of discrimination against each of the three different minorities is also statistically significant and there is also no significant difference in the discrimination rate for each of the three minorities. 9 So in answer to the second research question, hypothesis 2a is not supported: we find no significant variation in the extent of discrimination between the minority groups. This suggests that racial bias against non-white minorities is not playing a significant role in discrimination. These results support hypothesis 2b, that the discrimination rate is similar across groups. This is consistent with an explanation based on in-group favouritism.
Does this discrimination vary by occupation?
It could be that minority candidates are discriminated against when applying for some types of jobs but not others. To examine this, Table 3 provides relative callback rates and p-values arising from similar significance tests with respect to each of the three occupations involved in the experiment.
Callback rates and significance tests for discrimination against minority candidates by occupation
The similarity in the estimated relative callback rates reveals a consistent level of discrimination across each of the three occupations. Candidates with Irish names are over twice as likely to be invited to interview for all three occupations. Meanwhile the p-values in the final column confirm that this discrimination is statistically significant, albeit marginally so in the case of sales assistants, where the sample-size is smallest. Thus, the discrimination observed is not confined to a particular type of job, but applies across the three occupations involved in the experiment. Moreover, there are no significant differences in the extent of discrimination across the three occupations. 10
We cannot conclude that discrimination is higher in an occupation with high levels of interpersonal contact, like sales. So while the number of usable cases in sales is small, hypothesis 3a, derived from the ‘taste for discrimination’ approach, is not supported. Similarly, the prediction from hypothesis 3b, suggested by the statistical discrimination perspective, was that discrimination would be lower in accountancy, as qualifications are more specific. This is also not supported. The findings in answer to the third research question support the hypothesis that discrimination against non-Irish candidates does not vary across the occupations tested (hypothesis 3c).
It remains possible that an experiment with a larger sample size, different minorities, or a different set of occupations might detect significant differences by minority or by other characteristics related to the posts being applied for. However, these data reveal no difference in the incidence of discrimination by minority group or occupation. Instead, the results suggest that there is strong discrimination against non-Irish candidates that applies broadly across different jobs.
Discussion and conclusions
The answers to the three research questions are straightforward to summarize. Firstly, candidates with Irish names are over twice as likely to be invited to interview for advertised jobs as candidates with identifiably non-Irish names, even though both submit equivalent CVs. The chance of observing this outcome, if there were in fact no discrimination in the real world, is less than one in a million. Secondly, we find no differences in the degree of discrimination faced by candidates with Asian, African or German names. Thirdly, this finding does not vary across occupations applied for. The results indicate that there is strong discrimination against non-Irish candidates and this applies across a range of occupations in the Irish labour market. This is the first such field experiment in Ireland, and the first time discrimination in recruitment on the basis of ethnicity has been clearly demonstrated.
The findings from the experiment are broadly consistent with survey research which asks migrants themselves about their experience of discrimination in Ireland, though clearly the measurement is quite different. Russell et al. (2008) find that whereas 4.9 per cent of Irish nationals reported experiencing discrimination in recruitment in the last two years, 12.6 per cent of non-Irish nationals did. Where findings diverge is regarding national/ethnic differences in access to employment. O’Connell and McGinnity (2008) find that around 22 per cent of black respondents report recruitment discrimination, compared to 13 per cent of non-English speaking white respondents and 9 per cent of Asians. It could be that the perceived discrimination is due to other factors, like a period out of the labour market while an asylum applicant, though we cannot rule out that a significant difference would have appeared in our experiment between African and other minorities had the sample size been larger.
How does the discrimination rate compare with rates observed in international studies that use a similar methodology? In Ireland, the overall relative risk of callback in our study was just over 2. Considering other studies that test the first stage of the recruitment process using correspondence tests, relative callback rates vary considerably, from a low of 1.12 against Italians in Australia in 2007 to just over 2 against Africans in France in 2006 (Booth et al., 2010; Cediey and Foroni, 2008). Many studies find discrimination, on average, somewhere in the middle. Bertrand and Mullainathan (2004) testing in the USA in 2001–2 found that white Americans were 1.5 times as likely to be called to interview as African-Americans. Carlsson and Rooth (2007), testing in Sweden in 2005–6, find that Swedish applicants were 1.5 times as likely to get called back as those with a Middle Eastern name. The fact that discrimination recorded in the present experiment is high relative to similar studies carried out in other countries lends support to hypothesis 1b.
What are the theoretical implications of these findings? The findings are consistent both with explanations based on prejudice, at least a specific type of prejudice, and with those based on stereotypes. The discrimination rate is high, but then as discussed above, even relatively mild in-group favouritism could cause a lot of discrimination, if it is held by many people. In-group bias seems plausible in the Irish case, given a very short history of immigration, a cohesive national identity, and strong symbolic boundaries between different groups (Bail, 2008). In terms of differences between the groups, the similar discrimination rate between groups is also consistent with in-group bias, rather than prejudice directed at one particular minority group.
Equally, stereotypes that foreigners are inferior, either because of social skills or language skills, may have strongly framed employers’ reading of CVs. Or perhaps they did not even read past the name, and used it as a heuristic to infer poor language skills, in spite of the candidates having been educated in Ireland. The finding of little difference between the groups suggests that these stereotypes are directed towards a number of very different minority groups. Once again this is likely to be related to the recent nature of immigration to Ireland. There has been limited time for group-specific stereotypes to develop in Ireland, as employers have little experience of hiring minority groups.
Future research could investigate these explanations in greater depth. For example, a future experiment could vary the characteristics of the applicants, testing a high, medium and low-quality CV for Irish applicants, possibly compared to a high-quality CV for minority candidates, to explore what the ‘minority penalty’ is equivalent to in terms of qualifications and/or work experience. How much more qualified does a minority candidate need to be to achieve the same callback rate as an Irish candidate? This approach could also shed light on over-qualification, a well recognized phenomenon among migrants, in Ireland and internationally (Barrett et al., 2006; OECD, 2007).
A rather different avenue for future research would be to investigate the role of prejudice and stereotypes in recruitment discrimination using laboratory experiments (Krings and Olivares, 2007). These experiments, independent of the number of vacancies and the economic recession, could be used to explore the association between measures of in-group favouritism or stereotypes associated with different minorities and evaluations of CVs in Ireland. Alternatively a study could combine a field experiment in recruitment with a laboratory experiment using an implicit association test, like Rooth (2010). How do implicit attitudes influence recruitment discrimination?
The experimental method used in this article could be used in a wide range of settings, testing gender, family status, age, disability and social class, and has considerable appeal in research on labour market stratification. A key strength of experiments like this is that they can clearly measure discrimination on the basis of group membership, and in so doing measure the role of discrimination by employers in job allocation and social stratification more generally.
Our results imply that part of the reason that minority groups experience labour market disadvantage is because employers are not treating their CVs in the same way as identical CVs of Irish candidates, and they are thus not matched with jobs commensurate with their qualifications and experience. In the matching of individuals to jobs there is a ‘minority penalty’, i.e. an additional penalty for being a member of a minority group, and in this study, that penalty is considerable. Whatever the explanation for it, the high rate of discrimination in recruitment in Ireland is a cause for concern.
Footnotes
Acknowledgements
The experiment on which this article is based was funded by the Irish Equality Authority as part of a Research Programme on Equality and Discrimination, and we would like to acknowledge this funding. We would also like to thank our colleagues, Jacqueline Nelson and Emma Quinn, for all their work on the experiment.