Effecten van etniciteit bij de selectie van politieagenten
Engelse titel: Ethnicity Effects in Police Officer Selection: Applicant, Assessor, and Selection-Method Factors
Datum: september 2008
Auteur: dr. Lonneke de Meijer
Nederlandse samenvatting van Effecten van etniciteit bij de selectie van politieagenten: kandidaats-, beoordelaar-, en selectiemethodefactoren
De politie heeft de reputatie vooral mensen te selecteren die goed passen binnen de dominante, autochtone cultuur. Het personeelsselectieproces bij de politie is onderzocht, een organisatie die het van belang vindt dat er allochtone agenten in dienst zijn. Er werd bestudeerd in welke mate kenmerken van de kandidaat, de beoordelaar en het selectie-instrument bepalend zijn voor de bestaande scoreverschillen tussen autochtone en allochtone kandidaten. Verschillende zaken blijken van belang voor selectie in een multi-etnische context. Zo verklaart de beheersing van de Nederlandse taal een substantieel deel van de scoreverschillen tussen de autochtone en de allochtone groepen. Daarnaast vond de promovenda verschillen in de wijze waarop beoordelaars een eindadvies vormen. Voor autochtone en allochtone kandidaten werden scores op verschillende selectie-instrumenten op een verschillende manier gecombineerd in een eindadvies.
Beoordelaars voelen zich mogelijk meer onzeker bij het oordeel over allochtone kandidaten. Om dit te verbeteren moeten de beoordelaars meer ervaring opdoen bij het beoordelen van allochtone kandidaten, onderling kennis uitwisselen over het beoordelen in een multi-etnische context en een verdere standaardisatie van het selectieproces.
Een mogelijke oplossing om de verschillen in scores tussen autochtone en allochtone kandidaten bij personeelsselectie te verminderen, is de situationele inzichttest die werd ontwikkeld door de onderzoekster. Bij een situationele inzichttest zijn de scoreverschillen kleiner dan bij traditionele instrumenten, zoals de intelligentietest. Deze situationele inzichttest bestaat bijna volledig uit beeldmateriaal, er komt nauwelijks geschreven tekst in voor. De beheersing van de Nederlandse taal speelt dan een geringere rol.
Het onderzoek pleit voor meer standaardisatie bij de vorming van een eindadvies voor aannemen of afwijzen van kandidaten en het verbeteren van de taalvaardigheid van kandidaten die deelnemen aan het selectieproces.
De rest van deze tekst is in het Engels.
Purpose
Increase the knowledge about potential factors explaining differences between ethnic groups in personnel selection.
Research questions
1: Do Score Differences on Selection Measures between Ethnic Groups Exist?
2: Which Applicant Factors Play a Role in Ethnic Score Differences?
3: Does the Judgment Process Differ when Assessing Different Ethnic Groups?
4: Which Similarity Issues between Applicants and Assessors Play a Role in Score Differences?
5: What is the Predictive Validity of Cognitive and Non-Cognitive Measures?
6: Can Situational Judgment Tests Measure Integrity across Ethnic Groups?
Summary of Main Findings
In the introductory chapter, six research questions (RQs) were raised that directed the studies presented in this dissertation. These questions relate to he relative extent to which applicant (RQ1 and RQ2), assessor (RQ3 and RQ4), and selection-method factors (RQ5 and RQ6) are able to explain existing differences between ethnic groups.
Guided by these six questions, the main findings of the five empirical studies are outlined below, based on a large-scale dataset (more than 13,000 applicants) from the Dutch police officer selection procedure.
RQ 1: Do Score Differences on Selection Measures between Ethnic Groups Exist?
As a starting point for further research into possible explanatory factors for ethnic group differences in personnel selection, it was necessary to, firstly, determine whether ethnic differences in scores on selection measures exist. Score differences on the various selection measures (i.e., a cognitive ability test, a personality questionnaire, an Assessment Center (AC), an employment interview, and a final employment recommendation) were compared with each other. To this end, the four largest ethnic minority groups in The Netherlands each were compared to the Dutch majority group separately. In this comparison, first-generation minority groups were distinguished from second-generation minority groups. Significant score differences between the ethnic majority group and first generation ethnic minority groups to the advantage of the majority group existed on all selection tools. These differences were consistent with the findings from the literature. However, score differences between the majority group and the second-generation minority groups were much smaller, although the majority group still scored higher than the second-generation minority groups. The only exception to these score differences was the personality dimension Conscientiousness, on which minority groups systematically scored higher than the majority group. Score differences between the ethnic majority group and the ethnic minority groups were largest on the cognitive ability test and smallest on the personality questionnaire. Altogether, score differences on the personality questionnaire were rather unsystematic with sometimes the majority group and sometimes the minority group scoring higher. Score differences between the first-generation minority group and the second-generation minority group were the largest for the Antillean group, and the smallest for the Turkish group. Turkish minority applicants scored somewhat lower than the other ethnic minority groups on all selection measures.
RQ 2: Which Applicant Factors Play a Role in Ethnic Score Differences?
Investigating potentially explanatory applicant factors regarding the selection measure score differences, it was asked 1) whether applicant demographics (i.e., language proficiency, education, and ethnicity) were able to explain score differences on selection measures between ethnic groups, and 2) whether ethnic minority applicants whose demographics (i.e., language proficiency and education) were either very positive or very negative were evaluated more extremely (i.e., positive or negative, respectively) than ethnic majority applicants with the same very positive or negative demographics. The explanatory power was investigated of the applicant factors Dutch language-proficiency, education, and ethnicity on score differences on the so-called objective measures (i.e., the cognitive ability test and the personality questionnaire). Furthermore, regarding the so-called subjective measures (i.e., the AC, the employment interview, and the final employment recommendation), assumed-characteristics theory and complexity-extremity theory were tested. Score differences on the objective measures were explained mostly by the applicant factor Dutch language-proficiency. Both education and ethnicity explained only small proportions of the score variance. Supportive of assumed-characteristics theory, results showed that the applicant factors Dutch-language proficiency and education explained more of the variance in score differences on subjective measures than did the applicant factor ethnicity. In addition, this finding was more outspoken when assessors had knowledge about the applicants’ language-proficiency, education, and ethnicity compared to when they did not. Thus, having knowledge of someone’s demographic characteristics diminishes the influence of ethnicity per se, as assumed-characteristics theory would predict. In contrast to what complexity-extremity theory would predict, ethnic minority applicants whose Dutch language-proficiency and education were either excellent or very low were not evaluated more extremely (i.e., positive or negative, respectively) than ethnic majority applicants with the same demographics. Rather, the general pattern was that the ethnic minority groups scored somewhat lower on the subjective measures. Related to the effect of assessor factors on differences between ethnic groups of applicants, two research questions were formulated that guided the studies presented in Chapters 3 and 4. As assessor influence was the focus, these two studies concentrated on subjective measures solely.
RQ 3: Does the Judgment Process Differ when Assessing Different Ethnic Groups?
This question concerning assessor factors asked whether assessors integrate information into a final employment recommendation differently when judging an ethnic majority applicant than when judging an ethnic minority applicant. To this end a judgment-analysis study investigated assessors’ judgment processes evaluating ethnic minority applicants and their judgment processes in evaluating ethnic majority applicants. The sample only contained assessors belonging to the ethnic majority group. The term ‘judgment process’ is used to describe the process of giving weights to sources of information (i.e., scores on an AC, an employment interview, and a personality questionnaire) when combining these into a final employment recommendation. Thus, the focus is not on subgroup score differences on selection measures, but on differences in weights when combining information from various selection measures into a final advice.
Results showed that the assessors used more irrelevant sources of information and based their decision to a lesser extent on their own ratings than on ratings of others when judging ethnic minority applicants compared to ethnic majority applicants. Probably because of less experience or more uncertainty when making a judgment about ethnic minority applicants, assessors used more and also more irrelevant information in their judgment process. They were also hesitant to use their own ratings when assessing ethnic minority applicants and incorporated information from others more in their decisions about ethnic minority applicants compared to majority applicants.
RQ 4: Which Similarity Issues between Applicants and Assessors Play a Role in Score Differences?
The second question regarding the influence of assessor factors on differences between ethnic groups was directed at demographic and perceived similarity between assessor and applicant and at whether such similarity explains score differences between the ethnic majority group and ethnic minority groups. Using multilevel analyses, the effects are investigated of actual demographic similarity between assessor and applicant and of perceived similarity of applicants by assessors on score differences on subjective measures. Assessors came from both the ethnic majority and the ethnic minority group. From Social Identity Theory and perceived intergroup similarity, it may be expected that both demographic and perceived similarity will lead to higher evaluations. However, previous research had found mixed results. Most studies did not use multilevel analysis, while for some studies it would have been the most appropriate analysis technique given their nested data structure. In another study examined demographic similarity and analyzed their data with multilevel analyses. They found that demographic similarity between applicant and interviewer did not explain interview score differences between ethnic groups. The question, therefore, could be raised whether the same holds for other subjective measures and for perceived similarity. This study reported that neither demographic nor perceived similarity was able to explain score differences between the ethnic majority and the four ethnic minority groups on any of the subjective measures. Therefore, no evidence was found for (dis)similarity differentially affecting evaluations of ethnically diverse applicants during personnel selection.
The final two research questions focused on whether method factors, i.e., the psychological measures used, were able to explain differences between the ethnic majority group and ethnic minority groups.
RQ 5: What is the Predictive Validity of Cognitive and Non-Cognitive Measures?
The first question regarding method factors explored whether the predictive validity of the non-cognitive ability tests was comparable to the predictive validity of the cognitive ability test for both the ethnic majority and the ethnic minority group. As was the case for RQ3, the research guided by RQ5 took a somewhat different approach than the direct investigation of potential effects on score differences. The differential validity of various selection measures was examined. A distinction was made between the cognitive ability test and non-cognitive ability tests (i.e., the personality questionnaire, the AC, the employment interview, and the final employment recommendation). In contrast to the mostly U.S.-dominant literature on differential prediction until now, results indicated differential prediction on the cognitive ability test as well as on several non-cognitive ability tests, namely the AC, the employment interview, and the final employment recommendation. The cognitive ability test appeared to be more predictive of training success for ethnic minority trainees than for ethnic majority trainees. Yet, the AC, the interview, and the final recommendation were more predictive for ethnic majority trainees than for minority trainees. The personality questionnaire showed very little predictive power for both ethnic groups.
RQ 6: Can Situational Judgment Tests Measure Integrity across Ethnic Groups?
The second question concerning method factors studied to what extent a situational judgment test (SJT) developed to measure Integrity, indeed measured the same construct for the ethnic majority group and the ethnic minority group. In the final empirical chapter a study was described focusing on the construct-driven development of an Integrity-SJT, which was based on video scenarios. Although previous research had found that SJTs are often construct heterogeneous, the study found support for the construct validity of the Integrity-SJT for both ethnic groups. Furthermore, the results showed that the SJT score difference was substantially smaller than the score difference that is generally found on cognitive ability tests. The SJT, thus, appeared to be a useful measure of Integrity in a multi-cultural setting.
Conclusions
In conclusion, the research reported in the present dissertation has highlighted several issues. One important issue is language as Dutch language proficiency of applicants explained a substantial part of the score differences between the ethnic majority group and ethnic minority groups. Interestingly, assessor-applicant (dis)similarity did not differentially affect evaluations of ethnically diverse applicants. This finding alleviates concerns that discrimination of ethnic minority groups due to (dis)similarity may occur during personnel selection. However, a difference was found in the decision making process of ethnic majority assessors judging ethnic minority applicants compared to ethnic majority applicants. This finding indicates that assessors are, in some way, affected by the ethnicity of applicants. Gaining experience in assessing ethnic minority applicants, exchanging knowledge about assessment in a multi-cultural setting among assessors, or perhaps further standardizing the selection process should diminish differential effects. Furthermore, selection measures, both cognitive and non-cognitive, appear to differentially predict training performance of ethnic majority and minority trainees. A possible explanation of this differential effect may lay in the subjective evaluations of supervisors during training. Finally, scores on a newly developed situational judgment test (SJT) turned out to show substantially smaller ethnic group differences than generally are found on the cognitive ability test. These findings yield practical guidelines for personnel selection in a multi-cultural setting, such as further standardization of the decision-making process to hire or reject applicants and diminishing the influence of language skills of applicants by means of SJTs. More research is needed to further improve our understanding of personnel selection, specifically, and job opportunities, in general, in a multi-cultural setting.