Subjective performance evaluation and ethnic discrimination
in the Netherlands
Name: Juul Philippart Student number: 11140968
Thesis supervisor: dhr. prof. dr. V.S. Maas Date: 22 June 2017
Word count: 16,626
MSc Accountancy & Control, specialization Control
Statement of Originality
This document is written by student Juul Philippart who declares to take full responsibility for
the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This research examines the effect of having a Moroccan ethnic background on subjective performance evaluations in the Netherlands. Prior research has shown that having a non-western ethnic background can have a negative effect on performance evaluations in western countries. Some studies have shown that people with a Moroccan ethnic background have a lower average income and lower chances of getting a positive response on a job application in the Netherlands. However, most research has not been able to find evidence that ethnic background is the factor that causes these differences instead of other underlying factors. I hypothesize that having a Moroccan ethnic background will negatively affect subjective performance evaluations and that this effect is moderated by the evaluator’s attitude towards immigrants. I tested the hypotheses on three different performance levels using a case-based experiment. The experimental data do not support the hypothesis that having a Moroccan ethnic background will negatively affect subjective performance evaluations on any of the three performance levels. The experimental data found partial support for the hypothesis that the evaluator’s attitude towards immigrants moderates the effect of ethnic background on the performance evaluation.
Key words:
Contents
1 Introduction 6
2 Literature review and hypothesis development 9
2.1 Subjective performance evaluations 9
2.1.1 Psychology research on judgments and decision-making 10
2.1.2 Heuristics and biases 10
2.1.3 Role theory and social identity theory 11
2.1.4 Priming 12
2.2 Discrimination 12
2.2.1 Types of discrimination 13
2.2.2 Theories on the causes of ethnic discrimination 14
2.2.3 Ethnic discrimination worldwide and in the Netherlands 14 2.3 Effects of ethnicity and race on the subjective performance evaluation process 16
2.4 Hypotheses 19
2.4.1 Hypothesis 1 19
2.4.2 Hypothesis 2 20
2.5 Operationalization of variables 22
2.5.1 Ethnic background 22
2.5.2 Subjective Performance evaluation 22
2.5.3 Attitude towards immigrants 22
2.5.4 Performance level 23
2.6 Schematic overview of the research framework 23
3 Research method 24
3.1 Causal-model 24
3.2 Experimental design 24
3.3 Participants 25
3.5 The case 27
3.6 Subjective performance evaluation in the experiment 28
3.7 Manipulation of ethnic background 28
4 Results 30
4.1 Preliminary analyses 30
4.2 Hypotheses testing 37
4.2.1 Hypothesis 1 37
4.2.2 Hypothesis 2 40
5 Discussion and Conclusion 44
References 47
1 Introduction
This research examines if the ethnicity of an employee affects the judgments of an evaluator in subjective performance evaluations for different levels of performance. This is tested by priming the evaluator with an employee’s name that indicates a Moroccan ethnic background and measuring if the performance of this employee is evaluated differently by this evaluator.
Research on the factors that can influence subjectiveperformance evaluations is important because performance evaluations are an essential part of the management control system of almost every organization and in most cases, at least part of the performance evaluation is subjective, as opposed to objective. Meaning that the performance evaluation is at least partially based on the personal impressions, feelings, and opinions of the evaluator rather than calculated through formulas based on quantitative performance measures. The use of subjectivity can reduce the risk of misalignment for the performance evaluation system because it enables supervisors to provide a more accurate and complete picture of subordinates' performance than would be the case were their evaluations based solely on available objective performance indicators (e.g., Gibbs et al., 2004; Murphy and Cleveland, 1995).
Subjectivity, however, does not come without costs. Several studies have proven that many employees feel that subjective evaluations are sometimes inaccurate and unfair (e.g. Bol et al., 2016; Ittner et al., 2003). This may be due to the fact that subjectivity in the performance evaluation process allows for intentional and unintentional evaluation biases (Harris, 1994; Jawahar and Williams, 1997; Prendergast and Topel, 1993). Prendergast and Topel (1993) found that managers intentionally use subjective performance evaluations strategically to reward and promote their favorite subordinates or to advance their own position. Also, psychology research has shown that any type of information can be a prime and consciously and unconsciously influence decisions (e.g. Tversky and Kahneman, 1974).
One of the factors that cause the performance evaluation to become less fair and accurate are biases based on the ethnicity of the employee that is being evaluated. Over the years much research has provided evidence for economic ethnic and race discrimination (e.g. Becker, 1971; Greenhaus et al., 1990; Mount et al, 1997; Pendakur, 1998) and recent literature has shown that there are still significant differences in income and opportunities between ethnic groups (e.g. Andriessen et al., 2015; El Vira and Town, 2001; Lever and Waaijers, 2013). Andriessen et al. (2015) found evidence of ethnic access discrimination in the Netherlands. Specifically, they found that in The Hague applicants with a Moroccan name had a significantly lower chance of
getting a positive response on a job application compared to equally qualified candidates with a Dutch name. Lever and Waaijers (2013) found that non-western immigrants in the Netherlands have a significantly lower average income compared to native Dutch in the same age category, but that this difference is partly caused by the difference in education level and call upon Social Security between these groups and Veenman (2004) found that, conditional on having a job there is hardly any difference in wages and other job characteristics between second-generation immigrants and native Dutch of the same age group.
The purpose of this research is to contribute to the existing knowledge regarding subjective performance evaluations and ethnic discrimination in the Netherlands. It does so by investigating if the ethnic background of an employee can influence the performance evaluation decisions of the evaluator. The research specifically focuses on the influence of having a Moroccan name, as opposed to having a Dutch name, on the bonus that is subjectively assigned by an evaluator. The question that this research aims to answer is: Does ethnic background affect the subjective evaluation of performance?
The research question is relevant and important because, by understanding what the effect of ethnicity on subjective performance evaluations is and when the effect occurs, designers of performance evaluation systems will be better able to prevent ethnic discrimination in performance evaluations.
There is already a vast amount of research on both subjective performance evaluations and ethnic discrimination. By contrast, there is far less research that has combined these two research fields and investigated the effect of ethnic background on subjective performance evaluations. Also, most of this research has been inconclusive or contradictive. There are three main gaps in the existing research that this research aims to help breach. First of all, while these studies confirm that there is a difference in subjective performance evaluations among employees with different ethnic backgrounds and help explain how racial biases lead to outcome discrimination, they don’t include measures of actual employee performance. Therefore these studies cannot distinguish empirically whether evaluation biases are due to actual performance or to ethnic background or other factors. Secondly, most of the existing research was not done in the Netherlands. The differences between countries regarding ethnic discrimination (e.g. the ethnic groups that are being discriminated, the policies regarding ethnic discrimination, the types and severity of ethnic discrimination) make it difficult to generalize research findings across countries. Thirdly, most of the existing research is at least 10 years old. Because of the changing nature of issues regarding ethnic discrimination the results of these prior studies may not be
applicable to the current situation.
Overall, prior literature does not provide a clear consensus about whether ethnic background currently influences the subjective performance evaluation you get in the Netherlands. To my best knowledge, this is the first research that specifically examines the effect of having a Moroccan background on subjective performance evaluations in the Netherlands.
This research aims to clarify the relationship between ethnic background and subjective performance evaluations in the Netherlands by conducting an experiment that tests if evaluators give the same subjective performance evaluation to a manager with a Moroccan name as they do to a manager with a Dutch name, and if this effect is moderated by the evaluator’s attitude towards immigrants. It does so by manipulating the names of the managers that the participants evaluate while all other factors are held constant. Due to the design of the experiment, the research can provide evidence that the difference, if found and significant, is due to the manipulated factor, ethnic background and not due to other factors, because the other factors are the same in all conditions. It can also provide insight into the role of the evaluator’s attitude towards immigrants when making performance evaluation decisions. The participants in the experiment that provided the research data for this research are all currently living in the Netherlands and therefore this research provides specific insight in the effects of ethnicity in judgments and decision making in the Netherlands.
As a result, this research contributes to three different fields of research. First of all, it contributes to the existing management control research on performance evaluations and the factors that can intentionally and unintentionally influence subjective performance evaluations. Secondly, it contributes to the research on the effect of ethnicity on the performance evaluation process. And thirdly it contributes to the research that is determining the importance of ethnicity in the Dutch labor market.
The remainder of the paper is organized into 6 sections. Section 2 discusses prior research on subjective performance evaluation, decision-making, and discrimination and the hypotheses development. In section 3 the research method is described. The results are presented in section 4. Section 5 discusses the results, conclusions, and limitations of the research and gives recommendations for future research. Section 6 provides a list of references to the literature that was used in this research and section 7 contains the appendices.
2 Literature review and hypothesis development
This research builds on literature from three different fields. First of all, it makes use of the existing management control and psychology research on performance evaluations and the factors that can intentionally and unintentionally influence subjective performance evaluations. Secondly, it makes use of the current research on ethnic discrimination worldwide and ethnic discrimination in the Dutch labor market. And thirdly it uses the existing literature on the effect of ethnicity on the performance evaluation process. Based on the existing research in these three fields, the hypotheses are developed.
2.1 Subjective performance evaluations
Most organizations use some form of performance evaluations as one of their management control instruments. There are many different forms of performance evaluations but in most cases, at least part of the performance evaluation is subjective, as opposed to objective. In a subjective performance evaluation, the performance evaluation is at least partially based on personal impressions, feelings, and opinions rather than calculated through formulas based on quantitative performance measures. Subjective performance evaluations are non-verifiable, meaning that the correctness of a subjective assessment cannot be verified by a third party. There are several reasons for using subjectivity in performance evaluations. First of all, some actions cannot be adequately measured in an objective fashion while subjectivity allows evaluators to use any additional relevant information that arises during the measurement period. Secondly, subjectivity can mitigate potential incongruence effects such as a focus on short-term profit stopping managers from making necessary long-term investments. Third of all subjectivity can reduce risk imposed on managers by filtering out the impact of uncontrollable events (Baker et al., 1994; Baiman and Rajan, 1995).
Subjectivity, however, does not come without costs. First of all, subjectivity can lead to inaccurate performance assessments due to leniency and centrality bias, favoritism, and the personal impressions, feelings, and opinions of the evaluator. Secondly, subjectivity can cause subordinates to try to influence the supervisor inappropriately. Finally, subjectivity can lead to uncertainty about the measurement criteria among both the subordinates and the supervisors (Prendergast and Topel, 1993; Gibbs et al., 2003). There are different ways subjectivity can be used in a performance evaluation: (1) A performance evaluation can be fully or partially based on subjective judgments about performance; (2) the weights of some or all of the objective performance measures are determined subjectively; (3) the consequences of the performance
evaluation such as a bonus, promotion or dismissal are determined subjectively (Ittner et al., 2003; Gibbs et al., 2004). Often a combination of these different forms of subjectivity is used.
Although subjectivity can help to make performance evaluations better and fairer, it also allows for certain intentional or unintentional biases to influence the ratings given to employees. There is a lot of research on the different factors that influence judgments and decisions, such as subjective performance evaluations. This research is discussed in the next paragraphs.
2.1.1 Psychology research on judgments and decision-making
The performance evaluation process includes both judgments and decisions. A judgment is a comparison of a stimulus to another stimulus or the evaluation of a stimulus in relation to a standard (e.g. manager A's performance is better than manager B's performance or manager A's performance should be rated excellent according to the evaluation criteria). A decision is the choice of a stimulus (alternative, action) from a set of stimuli (Birnberg et al., 2007). How and how well individuals make judgments and decisions as well as the factors that can influence (economic) decision making have been researched extensively in psychology research (e.g. Baron, 2000; Hastie and Pennington, 1995; Kahneman, 2011; Tversky and Kahneman, 1974).
Psychology is the science of the human mind (e.g., attitudes, cognition, motivation) and behavior (actions, communications). Although other social science theories frequently used in management accounting research also aim to explain and predict behavior, psychology differs from them in focusing on individual behavior rather than organizational and social behavior and on subjective phenomena such as mental representations rather than objective phenomena such as market prices and quantities or organizational size and technology. Psychology includes many different fields, but the literature that is relevant for this research mainly relies on theories from cognitive and social psychology. Cognitive psychology is the study of psychological processes that influence thinking. These processes include attention, knowledge, judgments, decisions, and learning. Social psychology is concerned with how other people influence the minds and behavior of individuals and includes understanding people, attitudes and social influence, and social interaction and relationships (Birnberg et al., 2007, pp. 113-114).
2.1.2 Heuristics and biases
Several studies have examined how human information-processing capabilities and decision strategies influence the use of information when assessing performance. These studies suggest that cognitive biases can play a significant role in a subjective performance evaluation (Lipe and Salterio 2000, 2002).
In their research on probabilistic judgment Tversky and Kahneman (1974) explore the severe and systematic deviation between the judgments of individuals and the assumed judgment implied by optimizing models. Research consistently reports that individuals’ probabilistic judgments deviate systematically and severely from the judgments implied by these models. They identified cognitive processes called heuristics that can explain and predict these judgment biases. They find that the information-processing demands of strict optimization in complex tasks often exceed individuals' cognitive capabilities and that when this happens, they rely on heuristics. Tversky and Kahneman (1974) identify three heuristics that individuals use to subjectively assess and revise probabilities: availability, representativeness and anchoring and adjustment. Other research has identified many more additional heuristics but in this research, I only discuss availability and representativeness. Availability is the subjective estimation of the probability of an event by the ease with which instances of the event or similar events are brought to mind. An event is more available when it is more familiar, salient, recent, or imaginable. Representativeness is the subjective estimation of the probability that object A (sample) belongs to class B (population) by the degree to which A is similar to or resembles B. Probability estimates based on representativeness are not influenced by base rates, sample sizes, or regression to the mean (Birnberg et al, 2007). The third heuristic they identified, anchoring is not within the scope of this research and therefore left out of this review. The reason these heuristics lead to deviations between actual judgments and the assumed judgment implied by optimizing models is that often the availability and representativeness are not equal to the probability.
2.1.3 Role theory and social identity theory
Role theory uses a set of constructs derived from different fields to explain and predict how people function in a social context. The theory assumes that the behavior of individuals is influenced by role expectations and norms that are held by others about how individuals in a particular role are expected to behave (Deutsch and Krauss, 1965; Shaw and Costanzo, 1982).
Social identity theory is a part of role theory. Social identity theory assumes that individuals divide their social world into two groups: in-groups (the groups the individual is a part of) and out-groups (groups that the individual is not a part of). Individuals group themselves with others based on similarities during a self-categorization process. From this process rises their social identity. To what level an individual identifies him or herself with a group influences how they interact with members of that group, interpret information about that group and make decisions that affect that group (Lembke and Wilson, 1998). The more individuals socially identify with a group, the more they focus their effort on the group’s best
interest instead of their own best interest (Brewer, 1979), and the more likely they are to behave cooperatively when confronted with social dilemmas and increase their contributions of public goods to the group (Wit and Wilke, 1992).
2.1.4 Priming
There is a lot of psychology research that examines how exposure to a particular stimulus (a prime) can affect subsequent perception, judgment, and decisions. Primes can be explicit, implicit, or subliminal. Most research does not focus on economic decision making, but a few studies demonstrate that primes can influence how people play economic games (e.g. Kahneman, 2011; Tversky and Kahneman, 1979). Priming is thought to play a large part in the systems of stereotyping (Bargh et al, 1996). This process can be explained by automaticity, which is similar to representativeness and availability. If certain characteristics or trait descriptions are frequently used these descriptions can automatically be used when interpreting someone’s behavior without being aware of this. This can lead to behavior that is not in agreement with an individual’s personal beliefs (Bargh and Williams, 2006). Bargh et al. (1996) provide evidence that priming affects behavior even when the individual is not consciously aware of the priming stimulus.
In summary, subjectivity in performance evaluations allows for intentional and unintentional biases to influence the judgments and decisions of the evaluator in the performance evaluation process and therefore the evaluation of the employees. This causes the deviation between the judgments of evaluators and the assumed judgment implied by optimizing models. These biases can include favoritism based on the groups the evaluator identifies him of herself with. Priming, representativeness, and availability are thought to play a large part in this process and can consciously and unconsciously affect the behavior of individuals.
2.2 Discrimination
In this research I use the following definition of discrimination: Discrimination is the disadvantageous treatment of people because they belong to a certain group or are classified as such (Köbben, 1985; Veenman, 1990; 2004). This type of discrimination is based on group stereotyping. Group stereotyping is a situation in which there are commonly held beliefs in a population that members of a certain group may have some shared characteristics that affect their behavior or their abilities (Fershtman et al., 2005).
2.2.1 Types of discrimination
In this research, I make a distinction between different types of discrimination. The first distinction that is made is between ethnic discrimination, where there is a difference in treatment based on ethnic background and discrimination where there is a difference in treatment based on other factors such as race, gender, or age. Though ethnic discrimination and race discrimination are closely related, they are not the same. Ethnicity is based on the social and cultural groups someone belongs to, it refers to cultural factors such as nationality, culture, and language (Ethnic, 2017; Ethnicity, 2017). Race is determined by how someone looks, it refers to a person’s physical characteristics, such as skin color (Race, 2017; Racism, 2017). Consequently, there is a difference between discrimination based on race and discrimination based on ethnicity. But, there are a lot of similarities between these two types of discrimination making the literature regarding race discrimination relevant for my research. Therefore I included literature on race-based treatment discrimination as well as ethnic treatment discrimination in this literature review. Secondly, in this research, I focus on economic discrimination. When equal levels of productivity are not rewarded with equal levels of payment this is described as economic discrimination (Aigner and Cain, 1977, p. 177). In the labor market, this is the case when employees with equal levels of productivity do not receive the same amount of wages. Finally, this research focuses on treatment discrimination rather than access discrimination. Unlike access discrimination, which prevents members of a subgroup of the population from entering a job or an organization, treatment discrimination occurs when subgroup members receive fewer rewards, resources or opportunities on the job than they legitimately deserve on the basis of job-related criteria. Thus, this form of discrimination represents a situation in which the treatment of employees is based, at least partially, on their subgroup membership instead of their performance (Levitin et al., 1971). The types of discrimination used in this research are displayed in Figure 1.
Figure 1
Types of discrimination
Economic discrimination
Other types of discrimination e.g. race, gender, age Discrimination
Non-economic forms of discrimination Ethnic discrimination
2.2.2 Theories on the causes of ethnic discrimination
There are different theories that try to explain the underlying causes of discrimination. One theory is the taste-based approach by Gary Becker (1957). In his theory, treatment discrimination and access discrimination are explained by an employer’s disfavor to hiring employees from a specific group because they feel less connected to this group. According to Becker, the employer is more likely to hire employees from a group they feel more associated with because individuals have strong preferences for working with people they see as their equals. They are even willing to suffer financial consequences to avoid working with people from other groups. These financial consequences include hiring a less suitable candidate that is from the same group instead of a more suitable candidate from a different group (Berndt, 1991, pp. 180-181). The taste-based approach assumes the discrimination is done consciously to increase the percentage of people from the same group working in their team as much as possible.
The second theory that can help to explain ethnic treatment discrimination is statistical discrimination. This is a theory by Edmund Phelps (1972) and explains the wage inequality between ethnic groups through individuals being judged based on the average characteristics of the group they belong to rather than their individual characteristics. The information about the performance of employees, on which performance evaluations are based, is never complete. Therefore the judgments and decisions of the evaluator are not only based on the available performance information, but this information is supplemented with all other information available to the evaluator about the employees that are being evaluated. This information may include information about the ethnic group the employee belongs to and the characteristics the evaluator links to this group. This information will consciously and or unconsciously influence the performance evaluation the evaluator gives to the employee (Berndt, 1991, p. 182). Even
when these group characteristics are true as an average for the whole group they are inappropriate to use for individual performance evaluations. Nevertheless, if the average performance of a certain group is assumed to be different from the average of the population this influences the performance evaluation of a person from that group (Aigner and Cain, 1977, p. 179).
2.2.3 Ethnic discrimination worldwide and in the Netherlands
Many years of research on discrimination has provided much evidence for economic ethnic and race discrimination (e.g. Becker, 1971; Greenhaus et al., 1990; Mount et al., 1997; Pendakur, 1998). However, discrimination and the causes and effects related to discrimination are
constantly changing and therefore the results of research that was done twenty years ago may not be applicable to the current environment. Some more recent studies have shown that there are still significant differences in income and opportunities between ethnic groups (e.g. Andriessen et al., 2015; Elvira and Town, 2001; Lever and Waaijers, 2013). Mason (2010) and Fryer (2010) find that, even though it has severely reduced, a wage difference between black and white employees with the same level of skill still exists. They also find that the wage difference is largely due to the difference in education level between the groups. Lever and Waaijers (2013) have similar findings in their research about the wage difference between non-western immigrants and native Dutch in the Netherlands. They find that non-western immigrants have a significantly lower average income compared to native Dutch in the same age group. Their research also shows that part of the difference in wages is explained by the lower level of education and higher call upon Social Security of non-western immigrants compared to the native Dutch. In his research on second-generation immigrants in the Dutch labor market, Veenman (2004) finds that with respect to employment rates there are some differences across ethnic groups in the Netherlands, but conditional on having a job there is hardly any difference in wages and other job characteristics between second-generation immigrants and native Dutch of the same age group.
Andriessen et al. (2015) studied the role of ethnic discrimination of non-western immigrants in the labor market of The Hague. Specifically, they tested if having a Hindu or a Moroccan background has an effect on the chance of getting a positive response on a job application (e.g. getting invited for an interview) in The Hague. They tested this by applying for jobs using three fictive candidates that were equal except for their ethnic background. One of the candidates had a native Dutch background, one candidate a Moroccan-Dutch background and one candidate a Hindu-Dutch background. To manipulate the ethnic background of the candidates they used common Dutch, Hindu and Moroccan names for the candidates. Their research shows that having a Moroccan background has a significant negative effect on the chance of getting a positive response on a job application compared to a Dutch background. In the research they distinguish between two mechanisms, the psychological mechanism and the economic mechanism, that can cause the coherence between image and discrimination. They describe the psychological mechanism as the image about social and cultural differences which leads to seeing the non-western candidate as part of a different group. The economic mechanism is described as the image about lower productivity which leads to seeing the non-western candidate as a risk. To test whether the found access discrimination was due to the psychological
mechanism or the economic mechanism they repeated the experiment with different types of information added to the applications of the non-western candidates. There were two different conditions in this part of the experiment. In the first condition, information was added to the application of the non-western candidates to show their involvement in the Dutch community and identifying themselves as Dutch. In this condition, the chances of Hindu-Dutch candidates were equal to the native Dutch candidates, but the Moroccan-Dutch candidates still had significantly lower chances of getting a positive response. In the second condition, two years of work experience, relevant extra courses and a passage on motivation and devotion were added to the applications of the non-western candidates. In this condition, the chances of the Moroccan-Dutch candidates were equal to the, native Moroccan-Dutch candidates that did not have the extra work experience and courses. However, even with the added experience the Hindu-Dutch candidates still had significantly lower chances of getting a positive response on their application compared to the native Dutch candidates that did not have the extra work experience and courses. Andriessen et al. (2015) conclude by linking discrimination of Moroccan-Dutch candidates on the labor market of The Hague to the economic mechanism, because in their case additional work experience and relevant courses can compensate for the effect of ethnic background and linking the discrimination of Hindu-Dutch candidates to the psychological mechanism, because in their case the effect of ethnic background diminishes when their application shows involvement in the Dutch community.
To summarize, this research focuses on economic ethnic discrimination, but also includes relevant literature on race discrimination. There are different theories that try to explain how and why discrimination occurs such as the taste-based approach and statistical discrimination. These theories are based on psychology research on the factors that influence judgments and decision-making such as priming, representativeness and availability. With regards to economic ethnic discrimination in the Netherlands prior research is either inconclusive or contradictive. They find that there are differences between ethnic groups but if these are caused by the ethnicity or due to other factors is unclear.
2.3 Effects of ethnicity and race on the subjective performance evaluation process One of the factors that can cause the performance evaluation to become less fair and accurate are biases based on the ethnicity or race of the employee that is being evaluated. Several researchers have studied the effect of ethnicity or race on the performance evaluation process (Greenhaus et al., 1990; Mount et al., 1997; Elvira and Town, 2001). While these studies help
explain how racial and ethnic biases lead to outcome discrimination among employees of different races, they lack measures of actual employee performance. Therefore these studies cannot empirically identify whether evaluation biases are due to actual performance or to race, ethnicity, or other factors.
Elvira and Town (2001) examine the role of employee race, supervisor race and worker productivity on performance ratings within a large U.S. corporation. Their dataset contains information on job characteristics, employee characteristics, objective performance and subjective performance ratings. In their dataset, there are no salary differences between minority employees of different races and white employees when they control for job title and subjective performance ratings. This suggests that as long as the processes of performance evaluation and job allocation are unrelated to race there is no racial discrimination in this firm. To investigate if this is the case they test if the variation in subjective performance rating is explained by objective performance and if the variance that is not explained by objective performance is accounted for by race. Controlling for worker productivity and other demographic variables, they find that black employees receive lower ratings than white employees. Part of the remaining variance in performance evaluation ratings among racial groups is associated with the racial composition of the subordinate-supervisor pair. They find that white supervisors rate white employees significantly higher than black employees, and black employees also rate white employees significantly lower than black employees. They also find that these biases occur despite the availability of objective performance measures. They conclude that more work understanding the influence of race in the internal firm processes that determine promotions is welcome because minorities are more likely to have supervisors of a different race, differences exist in the performance-evaluation process, and ratings disproportionately disadvantage minority outcomes.
Greenhaus et al. (1990) examine the relationship between race, organizational experiences, job performance evaluations and career outcomes for black and white managers from three organizations. In their research, they examine both direct and indirect effects of race on job performance evaluations and career outcomes. They posit that race influences job performance evaluations through its effects on the organizational experiences but find limited support for this mediational process. They do find a modest but significant direct effect of race on performance evaluation ratings. Specifically, they find that compared to white managers black managers received lower ratings from their supervisors on their job performance and promotability. The research provides several possible explanations for this direct effect but due to the nature of the research, they cannot empirically distinguish the determinants of the effect.
Fang and Heywood (2010) study the wage difference in the Canadian labor market between visible minorities and other Canadians. The research is partially based on their previous research (Fang and Heywood, 2006) on the association between payment method and ethnic wage differentials in Canada. This research confirmed that ethnic minorities earn significantly less than non-minorities when the payment is time-rate based but receive essentially identical earnings in the output pay sector. The data they use shows that visible minorities earn 14,5 percent less than other Canadians. They examine the impact of ethnicity on earnings after accounting for the method of pay, minority language, and immigrant status. Their research demonstrates that the reported relationship between ethnicity and wage difference is largely absent for females. They also find that incorporating immigration and language as earning determinants tends to move ethnic wage differentials in favor of minorities.
In summary, several researchers have studied the effect of ethnicity and race on the performance evaluation process. This prior research indicates that ethnic and racial biases occur in performance evaluations and lead to significant differences in ratings based on subjective performance evaluations. The research however cannot empirically distinguish the determinants of the effect. An overview of the theory used this research is presented in Figure 2.
Figure 2
Overview of the theory
Social Identity Taste based Theory Approach
Direct relation Connected
Discrimination
Management Accounting
evaluations Research
Psychology Research Judgment and Decision-making
Priming Representativeness and Availability Statistical Subjective performance Cognitive biases Role theory Theories on Ethnic Discrimination
2.4 Hypotheses
Subjective performance evaluations are widely used because of the benefits they can provide compared to objective performance evaluations. Subjectivity allows for the evaluator to base the evaluation not only on quantitative performance measures and calculated through formulas but also include other factors in their judgment of the performance. Psychology research has repeatedly shown that people's judgments and decisions deviate from the judgments and decisions implied by optimizing models (e.g. Tversky and Kahneman, 1974). There are several factors that can cause subjective performance evaluations to deviate from what they are expected to be according to the optimized model and research on ethnic discrimination has shown that ethnicity is one of these factors.
2.4.1 Hypothesis 1
Prior research on whether, and if so how, performance evaluations are influenced by ethnic background and race has mostly examined whether evaluators are biased against certain ethnic groups (e.g. non-Western) or a certain race (e.g. brown or black skin color), such that they get less favorable ratings than western or white employees given similar levels of performance (e.g. Becker, 1971; Heywood and O’Halloran, 2005; Fang and Heywood, 2006). Elvira and Town (2001), Greenhaus et al. (1990) and Fang and Heywood (2006, 2011) have found evidence of such a bias in both Canada and America. They each found significant, though sometimes minor, negative effects of ethnic background and race on subjective performance evaluation.
A study on the relation between ethnic background and income in the Netherlands found that non-western immigrants have a significantly lower average income compared to native Dutch of the same age (Lever and Waaijers, 2013). But they also found that the difference in wages is partly explained by the lower level of education and higher call upon Social Security. Andriessen et al. (2015) provide evidence of ethnic access discrimination in the Netherlands. Specifically, they found a significant difference in the chance of success between Moroccan-Dutch applicants and equally qualified native Moroccan-Dutch applicants when applying for a job in the labor market of The Hague. The chances of getting a positive response were significantly lower for the Moroccan-Dutch applicants. Therefore I expect that having a Moroccan ethnic background will have a negative effect on subjective performance evaluation.
H1: Employees with a typical Moroccan name will get assigned a lower average bonus compared
Decision-making research has found that under different circumstances individuals use information differently. Therefore, the information on the ethnic background of an employee may be processed differently in the case of a high performing employee than in the case of a low performing employee and therefore have a different, for example, stronger or weaker, effect on the evaluator’s perception, judgment, and decisions. Accordingly, I expect that the influence of ethnic background on subjective performance evaluations may be different for different levels of performance of employees and I will test the effect for three different levels of performance.
H1.1: Employees with a typical Moroccan name will get assigned a lower average bonus
compared to employees with a typical Dutch name when their performance is average.
H1.2: Employees with a typical Moroccan name will get assigned a lower average bonus
compared to employees with a typical Dutch name when their performance is above average.
H1.3: Employees with a typical Moroccan name will get assigned a lower average bonus
compared to employees with a typical Dutch name when their performance is below average. 2.4.2 Hypothesis 2
Subjectivity in performance evaluations allows for intentional and unintentional biases to influence the ratings given to employees. Prendergast and Topel (1996) argue that subjectivity opens the door to favoritism, where supervisors act on personal preferences toward subordinates to favor some employees over others. Psychology research has shown that individuals divide their world into in-groups, being the groups that the individuals identifies him or herself with and out-groups. Brewer (1979) found that the more an individual socially identifies with a group the more they focus their effort on the group’s best interest instead of their own best interests and Berndt (1991) states that an employer may give better evaluations to employees that are from the (ethnic) group that the employer associates him- or herself with compared to employees from a different (ethnic) group even when their performance is the same. Therefore I expect that the attitude of the evaluator towards non-western immigrants in the Netherlands moderates the effect ethnic background has on the subjective performance evaluation given by the evaluator.
H2: Attitude towards immigrants moderates the effect of ethnic background on subjective
performance evaluation.
Based on decision-making research I expect that the moderation effect of Attitude towards immigrants on the effect of ethnic background on subjective performance evaluations
may be different for different levels of performance and I will test the effect for three different levels of performance.
H2.1: Attitude towards immigrants moderates the effect of ethnic background on subjective
performance evaluation when performance is average.
H2.2: Attitude towards immigrants moderates the effect of ethnic background on subjective
performance evaluation when performance is above average.
H2.3: Attitude towards immigrants moderates the effect of ethnic background on subjective
performance evaluation when performance is below average.
These hypotheses will test whether the same performance if evaluated differently if the name of the employee being evaluated is a typical Moroccan or a typical Dutch name and if this effect is moderated by the evaluator’s attitude towards immigrants.
2.5 Operationalization of variables
In order to test the hypotheses, I operationalize the concepts ethnic background, subjective performance evaluation, attitude towards immigrants, and performance level.
2.5.1 Ethnic background
To operationalize the independent variable ethnic background I use different names for the employees that are being evaluated, indicating different ethnic backgrounds. Based on prior research on ethnic discrimination in the Netherlands, indicating that Moroccan-Dutch have a lower average income and lower chances of getting a positive response on a job application compared to native Dutch, I chose to specifically test for the effect of having a Moroccan ethnic background. To make the evaluator aware of the ethnic background of the employees, they have either a typical Dutch or a typical Moroccan name. Which employee gets which name will differ among the three cases. The group where the employee has a typical Moroccan name will serve as the treatment group and the groups where the employees have Dutch names serve as the control groups. The manipulation of this variable will be explained more in the next section.
2.5.2 Subjective Performance evaluation
The effect on the dependent variable subjective performance evaluation will be measured through the average bonus that is assigned to the employees by the participants of the experiment. The bonus can be any amount between €0 and €10.000 and the participants have full discretion in assigning the bonuses. The reason for using a bonus as opposed to a non-financial rating is to make the participants aware of the importance of the evaluation decision. 2.5.3 Attitude towards immigrants
The second independent variable is Attitude towards immigrants, which is measured by the level of agreement of the participants with three statements that are presented in part two of the experiment. The following three statements were used to measure Attitude towards immigrants: 1. I prefer living in a neighborhood where most people are of a Dutch origin.
2. I believe having many immigrants of different origins is of great added value to the Netherlands.
3. All the different cultures in the Netherlands are a positive addition to the Dutch culture. Participants were asked to indicate their level of agreement with the three statements on a
five-Figure 3 Operationalization of variables Hypotheses 2.1, 2.2 and 2.3 Independent variable Performance level
↑
Moroccan or Dutch name
↓
Attitude towards immigrants
Independent variable
Hypothesis 2
→
Assigned bonusHypotheses 1.1, 1.2 and 1.3
Schematic overview of the theoretical model
→
Hypothesis 1↓
Performance evaluation Dependent variable↓
Ethnic backgroundpoint Likert-scale (1= strongly disagree, 2=disagree, 3=does not agree or disagree, 4=agree, 5= strongly agree). An average score was computed taking into account the negative wording of statement 1.
2.5.4 Performance level
The third independent variable is performance level. To operationalize performance level the case was constructed so that the three employees that are being evaluated each have a different performance level. In the case, the participants are introduced to three fictive employees, all in the position of sales manager. From the information in the case, it is made clear that Manager 2 has an above average performance, scoring highest on almost all performance indicators, Manager 1 has an average performance and Manager 3 has a below average performance, scoring lowest on almost all performance indicators.
2.6 Schematic overview of the research framework
Figure 3 gives a schematic overview of the research model, the operationalization of the
variables, and the hypotheses. The variable Attitude towards immigrants is included as a control variable in Hypothesis 1, Hypothesis 1.1, Hypothesis 1.2, and Hypothesis 1.3 and as a
3 Research method
This section presents the methods used in this research. The first part of the research consists of a literature review. Based on this review of the prior literature, the hypotheses are developed. In the second part of the research, these hypotheses are tested using the data collected through a case-based experiment. Experimental research is quite common in management accounting research (e.g. Bol et al., 2016; Maas and Torres, 2011), becauseexperimental research can help to answer questions that otherwise might have stayed unanswered (Keppel, 1973).
3.1 Causal-model
There are different types of causal models. In this research, I use an additive model. Additive models assume that the effect of a particular variable, in this case ethnic background, can be understood in isolation from other variables and other factors that might influence individuals’ minds and behavior. That is, they assume that the existence and magnitude of the effect are not conditional on the level of any other independent variable (Birnberg et al., 2007). Although the psychology theories used to develop the hypotheses specify a sequence of mental processes that produce the effects ethnic background can have on individuals’ minds and behavior, the model tests only the beginning and end of the sequence: ethnic background and assigned bonus. The intervening mental states and processes between seeing a name that connects someone to a specific ethnic background and the assignment of the bonuses are not tested by the model.
3.2 Experimental design
To test the hypotheses I used a case-based experiment with a 2 x 3 between subjects design. The design of the research instrument is modeled to the main aim of the experiment, which is to test the effect of ethnic background on subjective performance evaluation for different levels of performance. Besides the main research question, the design of the research instrument also allows testing for the influence of gender, age, and attitude towards immigrants. The task that the participants are asked to perform in the case and the structure of the case are based on the research instrument that was used by Bol et al. (2016).
In the experiment, participants were given a hypothetical case scenario in which they were asked to assume the role of regional manager and given the task to evaluate the performance of three sales managers. The case provided some general information on the company and the three sales managers they had to evaluate as well as some more specific information on their
performance over the past year. Based on this information they had to assign a bonus to each of the three sales managers.
I manipulated ethnic background by giving the three sales managers typical Moroccan and Dutch names and rotating them between the three managers. Each manager had a Moroccan name in one of the groups and a Dutch name in the two other groups. The name of the manager served as a prime, which can influence the behavior of the participants even without them being consciously aware of it. The exposure to the name that indicates a certain ethnic background could affect the subsequent perception, judgment, and decisions of the evaluator (Bargh et al., 1996). To test if the name of the manager did indeed influence the performance evaluation decisions of the evaluator I compare the bonuses that are allocated to the three sales managers in the three different groups.
3.3 Participants
In total 361 people voluntarily started on the experiment by clicking on the link that was sent out by email and through WhatsApp, Facebook, and LinkedIn. Of these 361 people, 165 completed the experiment. The final sample that provides the data for this research consists of 159 participants that responded to and finished the online survey and are currently living in the Netherlands. Because the aim of the research is to test the effect Ethnic Background has on subjective performance evaluation in the Netherlands I only used respondents that are currently living in the Netherlands. In Table 1 the demographic characteristics of the participants are presented. The average age of the participants is 35, which is relatively young for the role of a manager they were assumed to have in the experiment. There are slightly more male (55%) than female (44%) participants. The average education level of the participants is very high, which is not a true representation of the entire population but suits the role the participants were given in the experiment. The participants were randomly assigned to group 1, group 2 or group 3. Therefore the groups are approximately the same size and control variables such as age, gender and education do not differ significantly between the three groups. The demographics per group can be found in Table 4 in the results section.
The generalizability of the results of the research may be limited due to the method used for collecting the data. The participants are not a random sample, but all collected through my own network. Therefore the results and subsequent conclusions based on the data sample may not be generalizable to the population at large which limits the external validity of the research.
TABLE 1 Demographics
Participants Number of participants Percentage of participants
Group 1 * 54 34% Group 2 ** 54 34% Group 3 *** 51 32% Total 159 100% Age 17-20 16 10% 21-25 44 28% 26-30 18 11% 31-35 16 10% 36-40 10 6% 41-45 13 8% 46-50 11 7% 51-55 19 12% 56-60 10 6% 61-67 2 1% Gender Male 88 55% Female 69 43% Other 1 1% No response 1 1% Education level High school 16 10%
Community college (MBO) 6 4%
Bachelor of Applied Sciences 49 31%
Master of Applied Sciences 16 10%
Bachelor (WO) 12 8%
Master (WO) 60 38%
*Group 1: Manager 1: Niels Verheij, Manager 2: Michael van der Mersch, Manager 3: Soufyan Elya **Group 2: Manager 1: Soufyan Elya, Manager 2: Niels Verheij, Manager 3: Michael van der Mersch ***Group 3: Manager 1: Michael van der Mersch, Manager 2: Soufyan Elya, Manager 3: Niels Verheij
3.4 Procedures
The experiment was distributed by sending out an anonymous link through Facebook, LinkedIn, Instagram, WhatsApp, and email. The data collection started on May 31, 2017 and closed on June 7, 2017. The participants participated in the experiment through a link that randomly assigned them to one of the three cases. The experiment started with a short introduction making the participants aware that they were participating in an experiment, that they would remain anonymous, and that the information would only be used for this research. The first part of the experiment was a hypothetical case scenario ending with the task of allocating the bonuses. The second part consisted of a questionnaire that involved questions about demographics, a manipulation check, and statements regarding the participants’ thoughts on people with different ethnic backgrounds living in the Netherlands. After continuing to part 2 of the experiment the participants were not able to go back to the case scenario and the questions of part 1. The participants were made aware of this when deciding to continue to part 2. The case scenario and questionnaire can be found in the appendix.
3.5 The case
The case-based experiment that served as the main instrument for this research consists of two parts. Part one is a hypothetical case scenario about the Dutch division of an international chocolate and biscuits company, The Chocolate Factory (TCF). The company was described as an international company operating mainly in Europe and owning a market share of around 5% of the Dutch chocolate and biscuits market. The participants were given the role of the regional manager of TCF Nederland. In this role, they were given the task to evaluate the performance of the three sales managers working for TCF Nederland and assign a bonus to each of them. The participants had full discretion in assigning these bonuses as long as the bonus per sales manager was between €0 and €10.000. To make their decision they received two types of information about each of the sales managers. First of all, there was a short introduction paragraph for each manager including information regarding the manager’s personality, collegiality, and motivation. Secondly, the information included a list of the manager’s gender, yearly salary, market the manager is responsible for, sales revenues for this year, gross profit for this year, customer satisfaction (on a scale of 0-10), complaints, years working for the company, age, education, and new customers minus customers lost this year including the effect this had on the yearly revenues.
The case was constructed so that sales manager 2 has a relatively high performance, sales manager 1 has an average performance and sales manager 3 has a relatively low performance. The differences, however, are not huge. It was up to the participants to assess how they would reflect the difference in performance in the bonuses they assigned. After assigning a bonus to each of the sales managers the participants were asked to briefly explain the main reasons for assigning this amount to this sales manager. The reason for this was to enhance experimental realism and make sure the participants deliberately assigned the bonuses to the sales managers.
The gender of all three fictive managers in the experiment is male. The reason for testing the effect of ethnic background using male managers is because Fang and Heywood (2010) found that the effect of ethnicity was largely absent for females. Research by Andriessen et al (2015) provided evidence of ethnic access discrimination in the Netherlands. In their research, the biggest difference was found between applicants with a Dutch ethnic background and applicants with a Moroccan-Dutch ethnicity. Therefore I choose to specifically test the effect of male managers having a Moroccan ethnic background. Testing the effect for female managers and other ethnic backgrounds might be interesting topics for future research.
3.6 Subjective performance evaluation in the experiment
The dependent variable in this research is performance evaluation. Performance evaluations almost always have consequences for employees. Performance evaluations can translate into bonuses, increase in fixed salary, promotions, or a combination of those. To make the participants aware that their evaluation decisions had consequences for the sales managers I choose to link the evaluations directly to bonuses. I did this by asking the participants to assign a bonus between €0 and €10.000 to each of the three sales managers. Resulting in three dependent variables: Bonus 1, Bonus 2 and Bonus 3.
3.7 Manipulation of ethnic background
In the experiment, I manipulated the independent variable ethnic background by using Dutch names and a Moroccan name for the sales managers and rotating the names between the three sales managers in the three cases. The participants in the experiment randomly got assigned one of three cases. In all cases, the participants got exactly the same information about the case and the performance of the three employees was held constant in all three cases as well. Other than that the names of the employees rotate between the sales managers the three cases did not differ.
Table 2 gives an overview of the rotation of the names between the three cases.
Table 2
Manipulation of the independent variable
Group 1 Group 2 Group 3
Manager 1
Dutch name Moroccan name Dutch name Average performance
Manager 2
Dutch name Dutch name Moroccan name
Above average performance Manager 3
Moroccan name Dutch name Dutch name
4 Results
In this section, the results of the statistical analyses performed are presented. In the first paragraph, the preliminary analyses are presented to provide additional insight into the research data. In the second paragraph, the results of the hypotheses testing are presented.
4.1 Preliminary analyses
The preliminary analyses provides additional insight in the data used in this research. The participants in the sample are 159 voluntary participants, all living in the Netherlands at the time of the experiment. The demographics of the participants can be found in Table 1.
In Table 3 an overview of the variables used to test the hypotheses is given. The table provides a description and the way of measurement of the variables used. Most of the variables originate from the questionnaire filled out by the participants. In addition to these variables additional variables were created to be able to test the hypotheses. Attitude towards immigrants is measured based on the level of agreement with three statements. Statement one was worded negatively as opposed to statement two and three and therefore had to be recoded in order to calculate a score for the construct variable Attitude towards immigrants were a score closer to 1 refers to a negative attitude and a score closer to 5 refers to a positive attitude. To test if the
three items measure the same underlying factor I performed a factor extraction. The results of the factor extraction indicate that the three items measure the same underlying factor. This
resulted in using all three statements to construct the variable Attitude towards immigrants. The results of the factor extraction can be found in Table 21 and Table 22 in the appendix. The
average score of the participants is 3.53 and results of the analysis of variances indicates that the average does not significantly differ between the three groups (F= 2.288, Sig.= .105). The results of the analysis of variances can be found in Table 23 and Table 24 in the appendix. The variable group represents what group the participants got assigned to. This is either group 1, group 2 or group 3. Each sales manager had a Moroccan name in one of the three groups and a Dutch name in the other two groups. To be able to use the variable group to test the effect of ethnic background I computed three new variables, one for each sales manager. In the new groups, a participant got the value 0 if the sales manager they evaluated had a Dutch name and 1 if the manager had a Moroccan name.
Table 4 provides the descriptive statistics of the most important variables in totals and per group. The participants were randomly assigned to the three groups. This led to 54
participants in group 1, where manager 1 was named Niels Verheij, manager 2 was named Michael van der Mersch and manager 3 was named Soufyan Elya, 54 participants in group 2, where manager 1 was named Soufyan Elya, Manager 2 was named Niels Verheij and manager 3 was named Michael van der Mersch and 51 participants in group 3 where manager 1 was named Michael van der Mersch, manager 2 was named Soufyan Elya and manager 3 was named Niels Verheij. The difference in the names of the managers was the only difference between the cases that were presented to the three groups. An overview of the assignment of the names to the managers per group can be found in Table 20 in the appendix.
The sample means per group for the dependent variable Bonus 1 show that on average the average performing Manager 1 got assigned the highest bonus assigned when he was named Soufyan Elya (x̅= 5310.87) and the lowest when he was named Michael van der Mersch (x̅= 4942.78). The sample means per group for dependent variable Bonus 2 show that on average the above average performing Manager 2 also got assigned the highest bonus when he was named Soufyan Elya (x̅= 7345.90) and he got assigned the lowest bonus when he was named Niels Verheij (x̅= 6728.15). The sample means per group for dependent variable Bonus 3 show that on average the below average performing Manager 3 also got assigned the highest bonus when he was named Soufyan Elya (x̅= 3905.52) and he got assigned the lowest bonus when he was named Michael van der Mersch (x̅= 2945.24). Whether this difference is significant will be discussed in the following paragraph.
Table 3
Measurement of variables
Variable Description Measurement
Ethnic background In the case either manager 1, manager 2 1 = Niels, Michael, Soufyan (Group 1, 2 or 3) or manager 3 has a Moroccan name. 2 = Soufyan, Niels, Michael
3 = Michael, Soufyan, Niels
Ethnic background Manager 1 in the case had a Moroccan name 0 = Niels Verheij or Michael of manager 1 or did not have a Moroccan name. van der Mersch
1 = Soufyan Elya
Ethnic background Manager 2 in the case had a Moroccan name Same measurement as
of manager 2 or did not have a Moroccan name. Ethnic background of manager 1 Ethnic background Manager 3 in the case had a Moroccan name Same measurement as
of manager 3 or did not have a Moroccan name. Ethnic background of manager 1
Age Age of the participant in years Ratio
Gender Gender of the participant 1 = Male
0 = Female
0 = Other
Education Highest form of education the participant 1 = Primary school has finished or is currently receiving. 2 = High school
3 = Community college (MBO)
4 = Bachelor of Applied Sciences
5 = Master of Applied Sciences
6 = Bachelor (WO)
7 = Master (WO)
Attitude towards Average of the following three statements: 5-point Likert-scale immigrants 1. 'I Prefer living in a neighborhood 1 = strongly disagree where most people are of a Dutch origin' * 2 = disagree
2. 'I believe having many immigrants of 3 = neither agree nor disagree different origins is of great added value 4 = agree
to the Netherlands' 5 = strongly agree
3. 'All the different cultures in the *Before calculation the average, Netherlands are a positive addition to the scores of statement 1
the Dutch culture' have been reversed.
Ethnic background Agreement with the statement: 1 = True manipulation check 'All three managers were of a similar 2 = False
ethnic background'
Bonus 1 Annual bonus that is assigned by the Average amount assigned participants to average performing Manager 1 based on the given information.
Assigned bonus can be any amount
between €0 and €10.000
Bonus 2 Same as 'Bonus 1' but for above average
performing Manager 2 Average amount assigned Bonus 3 Same as 'Bonus 1' but for below average
Table 4
Descriptive statistics
N Mean
Std.
Deviation Minimum Maximum
Age Group 1 54 36.39 13.53 17 62 Group 2 54 32.54 12.79 17 67 Group 3 51 35.31 13.12 17 60 Total 159 34.74 13.17 17 67 Gender Group 1 54 1.46 0.54 1 3 Group 2 53 1.40 0.49 1 2 Group 3 51 1.49 0.50 1 2 Total 158 1.45 0.51 1 3 Education Group 1 54 4.93 1.62 2 7 Group 2 54 5.30 1.72 2 7 Group 3 51 5.22 1.81 2 7 Total 159 5.14 1.72 2 7
Attitude towards Group 1 54 3.36 0.84 1 5
immigrants Group 2 54 3.55 0.88 1 5
Group 3 51 3.71 0.77 2 5
Total 159 3.53 0.84 1 5
Ethnic background Group 1 54 1.91 0.29 1 2
manipulation check Group 2 53 1.91 0.30 1 2
Group 3 51 1.94 0.24 1 2 Total 158 1.92 0.28 1 2 Bonus 1* Group 1 54 5082.07 1758.60 32 8051 Average Group 2 54 5310.87 1837.40 1017 8480 performance Group 3 51 4942.78 1742.09 0 8023 Total 159 5115.10 1775.77 0 8480 Bonus 2** Group 1 54 7182.54 2064.99 2000 10000
Above average Group 2 54 6728.15 2205.67 39 10000
performance Group 3 51 7345.90 2034.51 1983 10000
Total 159 7080.62 2107.35 39 10000
Bonus 3*** Group 1 54 3905.52 2571.50 0 8667
Below average Group 2 54 2945.24 2101.03 0 8581
performance Group 3 51 3024.08 2180.08 0 7645
Total 159 3296.66 2322.89 0 8667
*Bonus 1: In group 1 Manager 1 was named Niels Verheij, in group 2 Manager 1 was named Soufyan Elya and in group 3 Manager 1 was named Michael van der Mersch
**Bonus 2: In group 1 Manager 2 was named Michael van der Mersch, in group 2 Manager 2 was named Niels Verheij and in group 3 Manager 2 was named Soufyan Elya
***Bonus 3: In group 1 Manager 3 was named Soufyan Elya, in group 2 Manager 3 was named Michael van der Mersch and in group 3 Manager 3 was named Niels Verheij
Table 5 gives an overview of the correlations for the most important variables, except for the variable Ethnic background which is displayed in Table 6. Bonus 1 is significantly positively (at the 0.01 level) correlated with Bonus 2 and Bonus 3. This indicates that participants that assigned an above average bonus to Manager 1 tend to also assign an above average bonus to Manager 2 and Manager 3 and participants that assign a below average bonus to Manager 1 tend to assign a below average bonus to Manager 2 and Manager 3. Gender, being male, is significantly negatively correlated with bonus 2. The other variables are not significantly correlated.
The qualitative categorical variable Ethnic background consists of three categories. In order to test the hypotheses the variable has been transformed into variables with two categories. The independent variables computed to measure the effect of ethnic background are Ethnic background of Manager 1, Ethnic background of Manager 2 and Ethnic background of Manager 3. In each variable a value of 0 indicates that the manager has a Dutch ethnic background and a value of 1 indicates that the manager has a Moroccan ethnic background. The correlation between the ethnic background of a manager and the assigned bonus is displayed in Table 6. The Pearson correlation indicates a significant positive correlation (Corr.= .189, Sig.= .017) between Ethnic background of Manager 3 and Bonus 3. The other variables are not significantly correlated.
In order to be able to test if the effect of ethnic background on subjective performance evaluation is different for different levels of performance, I first tested if the difference in performance between the three managers was indeed experienced as such by the participants. In Table 4 the average assigned bonuses are displayed. Bonus 1 shows an average assigned bonus of 5115.10 for average performing Manager 1, bonus 2 shows an average assigned bonus of 7080.62 for above average performing Manager 2 and bonus 3 shows an average assigned bonus of 3296.66 for below average performing Manager 3. To test if the perceived performance of Manager 1 is significantly different from the performance of Manager 2 and Manager 3 I analyzed the variances between the bonuses given to the three different managers. The results of the analysis of variances are presented in Table 7 and Table 8. The results indicate that the difference between the three managers is significant (F= 131.509, Sig.= .000). The difference in the perceived performance of the managers is confirmed by the statistics in Table 7 showing that the lower bound and upper bound for the 95% confidence interval of the mean do not overlap between the three managers.
