Discrimination on the labour market : is labour market racism mainly the result of racist preferences or uncertainty reduction through statistical generalisation?

(1)

Discrimination on the Labour Market

Is Labour Market Racism Mainly the Result of Racist Preferences or

Uncertainty Reduction through Statistical Generalisation?

Danique de Mol van Otterloo Student number: 10693866 Supervisor: dhr. dr. D. F. Damsma

Date: 04-‐01-‐2018

(2)

Statement of originality

This document is written by Danique de Mol van Otterloo, 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.

(3)

Introduction

According to the Universal Declaration of Human Rights article 23-‐2 everyone has the right to be equally paid for equal work. However, this is not always the case. Different groups receive different wages, probably because they differ in some characteristics valued differently on the market. According to standard economic theory, these differences in wages would arise because of difference in productivity. But practice shows this is not the only factor that causes differences in wages. Discrimination is the unequal treatment of people because they are thought to belong to an inferior group, based on ideology of inferiority. This also causes prejudices to arise, negative beliefs and attitudes towards racial outgroups (Williams, 1999). The fact that characteristics, such as gender, looks, sexual orientation, origin or race are valued differently on the market, while they are unrelated to productivity, indicates discrimination (Arrow, 1973). Economists are interested in this presence of discrimination in the labour market, especially since Becker and Arrow & Phelps introduced their discrimination models. The taste-‐based model of Becker (1957) and the statistical model of Arrow and Phelps (1972) have become the economic bases for preferences on the labour market, which arise from discriminatory attitudes (Guryan & Charles, 2013).

In The Economics of Discrimination, Becker described three models about employers, co-‐workers and customers, each of these with a different source of discriminatory taste. The theory says those with discriminatory tastes will sacrifice economic benefit in order to avoid interaction with a minority against which they discriminate (Becker, 1957). Arrow and Phelps on the other hand, believe the discrimination is because of limited information available about the productivity and characteristics of the applicant, such as a resume. By picking up as much signals as possible, the employer will try to estimate the productivity of the applicant. The less perfect the signal is, the more the employer has to rely on the generalizations of the productivity of groups consisting of people with the same characteristics. This way the employers will try to reduce the costs of information acquisition as much as possible during the hiring process. The generalization of the individual is used as a screening device. The imperfect information causes the employer to treat the individual applicant as a member of a group, which leads to discrimination (Arrow, 1973; Guryan & Charles, 2013).

When trying to enter the labour market, a person has to start applying for jobs. At this point, discrimination already occurs. When a job applicant sends a resume, he or she can already be discriminated by the presumed gender or race, based on the name. The outcomes of the field experiment by Bertrand &

Mullainathan (2004) show, according to names that they assume will be linked to a specific race, people with African American-‐sounding names will be more discriminated on the labour market. Last year, a student of the University of Amsterdam wrote her thesis about the influence of the subgroup of Islamic names as they were originally categorised as African-‐American names in the paper of Bertrand and Mullainathan. She found out that if the names were Islamic-‐sounding, they were more discriminated than the African American-‐ sounding names (Timmer, 2016). Which theory can explain the discrimination in this situation?

(5)

A lot of research is done describing the level of discrimination on the labour market in different situations. This paper will analyse part of these data from the perspective of the theories of Becker and Arrow & Phelps.

‘Is labour market racism mainly the result of racist preferences or uncertainty reduction through statistical generalisation?’

The different forms of discrimination will be discussed first. Secondly, we will look at methods to test discrimination. Then we will discuss the fact that the level of discrimination changes over time. After, the relevance of a name towards the socioeconomic status will be discussed. And finally we will look at result of previous research and try to interpret those according to the theories of Becker and Arrow & Phelps.

Forms of discrimination

There exist large gaps between the earnings of people from different groups. This does not only vary between races, but also between genders. White men in the United States for example, earn $55,900 in 2006, whereas white women earn $35,000. This means the average white man earns 59,25 percent more than the average white woman (Borjas, 2010).

According to Borjas (2010) difference in education is a factor why wages may differ between groups, for example between people of different race. For men in the United States in 2007, 15 percent of white men, almost 20 percent of black men and over 40 percent of Hispanic men did not graduate high school. 29.9 percent of white men, 18.0 percent of black men and 11.8 percent of Hispanic men did obtain a bachelor’s degree or more. This could mean the actual productivity of these people is lower, which will cause lower wages. But these figures will also contribute to the general picture people have of different ethnic groups. When seeing an individual, people will generalize and make particular assumptions reflecting this general picture. For example, when seeing a Hispanic man, someone could think he will be less educated than when he sees a white man. Then that particular individual is not actually less productive, but just discriminated and thought to be less productive.

This latter is a form of discrimination that is called statistical discrimination, as Arrow and Phelps described it. This type of discrimination can always occur, even to a person who is not prejudiced at any point. Let’s assume, for example, an employer is colour-‐blind, gender-‐blind and profit maximizing. If two applicants apply for the same job, and the two applicants are exactly identical (race, age, education, working experience), except for gender. The employer still needs to make a decision somehow. Therefore, he will check the statistical information available about men and women related to employment. This way, the employer tries to take away part of the uncertainty that arises because of the imperfect information about the true productivity of the two applicants according to their resumes and application process. He will rely on the statistics about the average performance of the group. Consequently, the employer could hire the man instead of the woman, because figures show women quit jobs earlier than men (Borjas 2010).

According to Becker’s taste based model of discrimination, there are several forms of discrimination. When an employer is prejudiced and dislikes a

(6)

specific minority, the employer actually gets disutility when hiring someone of this minority. The prejudice may be reflected in the wage; the perceived costs of the employer may exceed the actual cost of hiring the employee. It also exists the other way around. Then it is not called discrimination but nepotism; you do not dislike, but prefer a specific person. This way the employer will act as if it is cheaper to hire this preferred person than it actually is. Discrimination (or nepotism) does not only exist with employers related to hiring employees, but it also exists between employees. If an employee dislikes working with a specific minority, the employee’s perceived wage will be lower than the actual wage. Also, if a customer dislikes buying a product from a particular person, the perceived price of the product will be higher than the actual price. The discriminating person will determine what type of discrimination is concerned: employer discrimination, employee discrimination or customer discrimination (Borjas, 2010).

In this paper, results of various articles will be compared to look what model or theory matches the type of discrimination. If there is difference in the level of discrimination when the only difference is race, this may indicate discrimination based on taste. It is likely that no other factor can be assigned to cause the discrimination in such a case, when all other factors are equal.

For the statistical discrimination it is possible to check if the difference in the level of discrimination changes by changing one factor such as education, neighbourhood or the skills required for the job. If there is a changing level of discrimination, this may be caused because of the different levels of statistical information that is used during the hiring process.

However, first we will discuss different methods used for testing discrimination in the next chapter.

Methods to test discrimination

It is difficult to test to what degree employers discriminate. Nevertheless, there are several ways to try to test the discrimination during the hiring process in the labour market. One way of testing discrimination is to do a correspondence test. Correspondence studies create fictitious resumes that are sent to real job offers by mail or fax. The resumes will be as realistic as possible and the race will be assigned randomly. The results are measured by the amount of callbacks or mails of the employer to fictitious phone numbers and mailing addresses (Pager, 2007). With an audit study it is impossible to assign race randomly to a real person, but with a correspondence study it is possible to use the fact the employer will make assumptions about the race of the job applicant based on its name. By choosing distinctive names you can randomly assign the race the employer perceives the job applicant to be (Guryan & Charles, 2013). However, with the correspondence test, you can only include a limited sample of jobs and there will be problems with signalling key applicant characteristics due to the fact that this test relies on paper applications only (Pager, 2007). Examples of correspondence tests looking at racial discrimination on the labour market are those of Carlsson & Rooth (2007) and Bertrand & Mullainathan (2004).

(7)

A different approach is an audit study. An audit study is a field experiment that has a few similar aspects with respect to the correspondence test, because it also uses fictitious résumés. However, in this case, the résumés are randomly assigned to white actors and comparable minority actors. Two of these actors with almost the exact same characteristics, such as weight, age and height are matched in pairs and sent to real job interviews. The actors will be trained to react identical to their assigned profile during the job interview. This way the only difference will be the single characteristic the researchers want to investigate and no other characteristics that the employer will take into account while making hiring decisions (Gaddis, 2014). However, one of the weaknesses it that it is difficult to make sure every single characteristic that may affect the employer in his hiring decisions is identical between the matched actors. When this is not the case, it will bias the outcome of the study (Pager, 2007). Examples of audit studies looking at racial discrimination on the labour market are the studies from Gaddis (2014) and Pager, Bonikowski & Western (2009).

To give an example of how a correspondence study works, one example will be described. One of the most widely cited papers investigating the effect of having a distinctive Black-‐sounding names instead of a White-‐sounding name during the hiring process is the paper of Bertrand and Mullainathan (2004). In this paper they investigated discrimination on the labour market with a large field experiment, a correspondence test. Bertrand and Mullainathan constructed over 4,000 fictitious resumes. Half of the resumes had an African-‐American-‐ sounding name, whereas the other half had a White-‐sounding name, all randomly assigned. They chose to vary credentials of the resumes, experimentally. This way they could send four resumes per job advertisement; 2 high quality resumes and 2 low quality resumes. The quality of the resume was determined by labour market experience, the career profile of the job applicant, the existence of gaps in employment, and skills listed. Responding to approximately 1,300 job advertisements in Boston and Chicago, almost 5,000 resumes of fictitious job applicants had to be constructed. The names were chosen according to the frequency data of all children born in Massachusetts in the period 1974-‐1979. Using ratios, they distinguished which names could be used as African-‐American-‐sounding or White-‐sounding names and chose one for each race: 9 African-‐American-‐sounding and 9 White-‐sounding names. As a check, they put these names in a survey and asked the respondents to identify the name as ‘White’, ‘African-‐American’, ‘Other’ or ‘Cannot Tell’. In the period of July 2001 until January 2002, the experiment was carried out in Boston. In the period of July 2001 until May 2002, the experiment was carried out in Chicago. Bertrand and Mullainathan estimated the marginal effect of the resume characteristics on the likelihood of getting a callback for an interview after sending the resume. They concluded job applicants with a White-‐sounding name receive 50 percent more callbacks for interviews compared to African-‐American names.

If they Bertrand and Mullainathan wanted to do an audit study, the résumé selection probably would have been the same, but they would have randomly assigned those to white and minority actors who would actually go to a job interview.

(8)

Both of these tests will try to simplify the situation as much as possible and try to find a cause for the discrimination. However, the different forms and causes of discrimination are difficult to distinguish. Most of the time there are several arguments pro and against a particular theory or model. When looking at the models of Becker and Arrow & Phelps, it is hard to state which model is the most accurate or explains all situations in which discrimination occurs, why it occurs. Which model is the best? When looking at results of previous research it may be possible to find some difference in causes. If discrimination happens most, for example, in jobs that include complex tasks, it would be reasonable to say that imperfect information causes the discrimination because the employer will generalize the applicant or employee more frequently based on statistical information. In that case the model of Arrow & Phelps would be more accurate. On the other hand, if discrimination occurs just as much in jobs with very simple tasks for which you will not need explicit or difficult qualifications, it could be stated that the discrimination happens based on only preference with the taste-‐ based model of Becker.

Next, we will look at the results of previous research and look at the interpretation of those findings.

Previous research

Pager, Bonikowski & Western (2009) show results of an audit study with black, white and Latino job applicants. The study consisted of applications to 171 employers in the low-‐wage labour market in New York. They tried to match the testers as perfect as possible; the testers were assigned fictitious résumés so they had identical abilities, the same level of education and work experience, and the same neighbourhood of residence. According to the hypothesis above, all else equal, there should be no difference in callbacks between the three different races. However, there was a clear difference in callbacks, even though they had these same characteristics. Only 15.2 percent of the black applicants received a callback or job offer, whereas 31.0 percent of the white applicants received a callback or job offer and 25.2 percent of the Latino applicants. Results even show that white applicants, who just got out of jail, had similar results as black applicants with clean backgrounds.

These results show that Becker’s theory may apply in this article; even though the résumés had the same characteristics, there is a clear difference in callback rates or job offers between these races. This study of Pager, Bonikowski & Western (2009) wants to point out the systematic forms of racial discrimination. It is not possible to state the difference in callback rates is completely due to discrimination based on preference, but the results suggest that a certain part of it is caused by preference-‐based discrimination. They even show there is a racial hierarchy visible; blacks at the bottom, whites at the top and Latinos in between, which was also mentioned by Bonilla-‐Silva (1997).

Bertrand & Mullainathan (2004) too looked at the difference in callback rates, but they also looked at the variation of the callback rates between different occupations. However, they did not find any variation in the racial gap. The callback rates did not differ across occupation categories or jobs with different skill requirement; the employers significantly prefer white applicants to African

(9)

American applicants, even in the better jobs. This suggests this racial gap is due to preference-‐based discrimination.

Bertrand & Mullainathan (2004) also did research on the effect of the quality of the résumés of the applicants. The difference in callback rates between white applicants with a high-‐quality résumé and a low-‐quality résumé was 27 percent (2.29 percentage point). For the African American applicants there was much less of an increase in callback with a difference of only 8 percent (0.51 percentage point). According to the findings of this experiment, returns on the different characteristics of the résumé are statistically and economically weaker for African Americans than for white applicants. This indicates African American employees profit less from investing in the quality of their résumé. However, assuming that discrimination is not preference-‐based, you would think the increase in quality of a résumé would have the same effect on both African Americans and white people. Since this is not the case, it indicates discrimination based on preference.

Kelman (1991) discusses the probability of existence of some form of systematic discrimination due to differentiation in the low-‐wage or low-‐skilled occupations causing differences in callback rates. Grodsky & Pager (2001) write about the different dimensions of growing importance interpersonal and manual skills represent. Due to the article of Moss & Tilly (1996), where they argue that communication skills and personality of black people are devaluated by employers relative to white people, Grodsky and Pager (2001) expect this devaluation might cause greater inequality between black and white people in low-‐earning occupations. This is because interpersonal skills characterize these low earning occupations. Also, they argue that black people might differentiate along the lines of occupations requiring manual skills, including physical strength and finesse. These skills may be rewarded when interpersonal skills may not. And because employers devalue those skills for black people, this may result in black people being encouraged to enter occupations with an emphasis on manual skills.

However, in 2015 Pager & Pedulla find there are only a handful of occupations showing racial differentiation, in contrast to the expectations above. Also, the few significant numbers for racial differentiation only show small effect sizes. Furthermore, Pager (2015) finds black people are less likely to apply for construction occupations, requiring manual skills, and are more likely to apply for health care support, office support, protective services, community and social services and sales. This is in contrast with the hypothesis of Grodsky & Pager (2001) stating black people may differentiate along lines of manual skills.

Carlsson & Rooth (2007) did a field experiment similar to the experiment of Bertrand and Mullainathan (2004). Carlsson & Rooth (2007) sent equal applications to 1552 different job openings in Sweden. The only difference between the applications was the name: either a native-‐sounding name or a foreign-‐sounding name, in this case Swedish or Middle Eastern names. Looking at the results of the experiment, callback rates for applications with a Middle Eastern name are fifty percent lower than callback rates for those with a Swedish-‐sounding name. However, in this field experiment there is a clear difference in the callback rates relative to the level of skills required for the job.

(10)

The high-‐skilled jobs, such as computer professionals, do not show a significant difference in relative callback rates. In contrast to jobs such as restaurant workers or motor vehicle drives, where the relative callback rate is the highest. This indicates a higher level of unequal treatment in low-‐skilled jobs relative to high-‐skilled job,s and thus a negative relation between the skill requirement and the relative callback rate for the job. In this case, because there is no significant difference in callback rates in high-‐skilled jobs, this might also suggest that it is because of the highly educated employers that there is less discrimination (Carlsson & Rooth, 2007). Apostle, Glock, Piazza & Suelzle (1983) show the more educated people are, the more likely it is they will reject negative racial stereotyping. If the level of education influences the degree of discrimination, this might affect the choice of the model being used. With complex jobs, you can assume the employee is highly educated. This means the employee is less discriminative based on preference and the discrimination that may occur will probably be due to statistical discrimination. The other way around, there will be poorly educated employees in simple jobs, so they might be more discriminative based on preference. If this is the case in the experiment of Carlsson & Rooth (2007), than it would mean the employees of the high-‐skilled jobs do not discriminate based on preference, but the little discrimination that still occurs may happen based on statistical information.

These results however, can also be explained from the point of view from Becker’s model. In the low-‐skilled jobs, such as a restaurant worker, there is frequent customer contact, so the highest relative callbacks in these jobs are in line with the customer discrimination of Becker. As a computer professional you will have less contact with customers and indeed it shows no difference in relative callback rates. However, contradicting to this, accountants and teachers in math and science are valued as high-‐skilled jobs and also show no difference in callback rates, but do have frequent customer contact, even if of a different type.

One would think the argument of education works both ways, according to the theories discussed above; high-‐skilled jobs require more information of the employer so this way statistical information might increase due to imperfect information and thus statistical generalisation, and high-‐skilled jobs include highly educated employees so there should be less discriminative taste or preference. Unfortunately, this is not applicable in every situation and does not explain all discrimination. This is also the case for the frequent customer contact as shown above. It explains parts of the discrimination, but not all.

The influence of the person who is hiring should not be underestimated. Stoll, Rahpael & Holzer (2004) and Giuliano, Levine & Leonard (2009) find black managers hire more blacks than non-‐black managers do, and less whites. Stoll, Raphael & Holzer (2004) their results show when the hiring manager is black, the black applicant has a 20 percent higher probability of being hired.

These results can be explained in the light of both theories. It can be explained from Becker’s point of view due to different tastes. A black hiring officer is likely to like black people more than a white hiring officer. On the other hand, in the light of Arrow & Phelps’ their theory, a black hiring officer is more familiar with black people and therefore the use of statistical information is different due to decline of the level of stereotyping and generalization.

(11)

Interpretation

The different results of previous research show there is not just one clear model to certify every discriminative situation in the labour market. Another possible way to think about models, is according to the theory Rodrik describes in his ‘Economics Rules: The Rights and Wrongs of the Dismal Science’ (2015). In the economic world, there is not just one single simplistic specific model. Economists create a lot of different models all together. Next, they try to link all those models to each other and map between the models and the real world. This way, economists try to reflect the diversity and the complexity of the real world, trying to explain everything. There is one pitfall however; economists tend to misuse the models because they assume that a model is the model, which can be used under all conditions. But according to Rodrik (2015) there is a minimalistic chance that a universal, general-‐purpose model will be developed. This is because every social setting needs a different model. When the setting, or circumstances, change, the economist needs to reconsider the model he uses and may even need to change to a different model. So maybe there is not just one best model, but in different settings or circumstances, different models are best. For example, in the article of Carlsson & Rooth (2007) they also added other data; register and interview information. There it shows when a female is responsible for deciding whom to call back, the relative callback rates will be lower than if a male is responsible. They also suggest companies with a higher personnel turnover minimize statistical discrimination due to the fact that they might have a more comprehensive recruitment process or have a lower threshold for calling applicants for an interview. These are even more factors to take into account when looking for causes for discrimination on the labour market. For example, the growth or neighbourhood of the company, the number of employees working at the company, the male-‐female ratio, the hiring process or different combinations of those factors. It differs in every situation to what extend there may be discriminated and based on what cause, model or theory.

Another explanation for the different results of previous research is because in every situation the level of taste or prejudice may be different. Segregation, for example, could influence the level of tastes. When the black/white-‐ratio is very high, white people might feel intimidated or threatened and therefore increase their prejudice. On the other hand, this could also ensure decreasing prejudice due to the more frequent contact with black people. The frequent contact and familiarity may decrease the level of misunderstanding and stereotyping. Also, this decreasing level of stereotyping could change the level of statistical discrimination.

Such changes in level of taste or discrimination over time are important to take into account. As a consequence, in every different situation or market, there are different levels of taste or discrimination. This makes testing discrimination across markets questionable (Cain, 1986). In the next chapter we will discuss historical events or other trends that may have an impact on changing attitudes towards minorities and the level of discrimination. This is necessary to take into account because this may bias the results.

(12)

Changing levels of discrimination

Terrorism

An interesting aspect of the paper of Betrand and Mullainathan is the fact that the 9/11 attacks were within the timeframe of the experiment. This event could have biased the outcome of the research.

According to the Toronto Police Service Hate Crime Unit, a 66 percent increase of hostile acts was reported in 2001. A large part of this increase of hostile acts was assigned to being a consequence of the terrorist attack on 9/11. 121 of the 338 hostile acts happened in September and October 2001. Almost half of the acts were targeted against Muslims, whereas only one hate crime against a Muslim was recorded in 2000 (Helly, 2004). The increased violence against Muslims could have caused an increase in the discrimination on the labour market. Even though Betrand & Mullainathan (2004) used Black-‐sounding names, a terrorist attack that increased the discrimination of Muslims still could have biased the outcome because of the distinctive names they chose in the experiment since some of those names may seem Islamic instead of African-‐American. This is an example of how preference or taste may change over time, due to an event or a different cause, and therefore change the level of discrimination.

The Black Power movement

This event mostly influences the level of discrimination of black people. Until the 1960s, white and black people had very similar names. However, this changed in the early 70’s; black people started choosing more distinctive names for their children. Fryer and Levitt (2004) stated this was consistent with a model that combines the rise Black Power movement and how black people thought about their identity. People believed you should be proud to be black. Giving your child a name that was not similar to a name a white child would get could be an example. The goal of the Black Power movement was to achieve black empowerment. Some thought of the Black Power as a way to achieve more equality and elevate the status of African Americans in society. On the other hand, there were also more radical people who thought of the Black Power as Black Nationalism and did not want to integrate with the white society (Patterson, 1996). There were several organizations, where there were two most significant: the Black Panther Party and the Nation of Islam (NOI). Especially the NOI could have had an impact on changes in names of black people. Elijah Muhammad was the African-‐American religious leader of the NOI after founder Wallace Fard Muhammad disappeared in 1934. He spoke about the Islam as a hidden religion, a hidden history. He told his followers about Allah, who according to him really was an Asiatic black man – the Original Man and the image of God. Allah was the founder of civilization and the colour of humanity was originally black. A few were genetically manipulated by a genius called Yacub who lived more than 6.000 years ago, that is how white people were created. Yacub made his own colony of white people who were incapable of doing good (Ogbar, 2005). This was the origin of the black people according to the NOI. Believes of the NOI were based on believes of the Islam, although they differ fundamentally. For example, the Islam believes Allah created mankind from a single man and female, so every human is in a way related to all other humans. This is different from the NOI’s theory about the evil colony Yacub

(13)

created 6.000 years ago and that this is the reason why white people are actually inferior to all other races. Nevertheless, the members of the NOI call themselves Black Muslims.

In the United States, the Federal Bureau of Investigation reported a 1,600 percent increase in hate crimes against Muslims in 2001. According to the reported crimes of the Toronto Police Service Hate Crime Unit, 57 hate crimes against Muslims occurred in September and October 2001 compared to just 1 hate crime reported in 2000. Because the 9/11 attacks were organised by the Islamic terrorist group Al Qaeda, the increase in these hate crimes against Muslims are thought to be related to the terrorist attacks. In August 2002, a survey in Canada showed 33 to 45 percent of several cities admitted that the 9/11 attacks made them mistrust Muslims more (Helly, 2004).

These events can be an extra explanation for the results of the experiment of Bertrand & Mullainathan (2004). A large part of people looking for a job in 2001-‐2002 is born after the 70’s, the start of the Black Power movement. This also means there might have been influence of the Nation of Islam on these people. The strong Black Nationalistic influence of the NOI may have resulted in some distinctive African-‐American names, which are also distinctive for Muslims because of the (semi-‐) Islamic origin of the NOI. This is why the experiment of Betrand & Mullainathan (2004) for example, may overestimate the discrimination of African-‐American people by names associated with the Islam as Timmer (2016) concluded.

From Arab to Muslim

Widner & Chicoine (2011) show there is an increase in discrimination of Arab-‐American people after 9/11. However, Arabs were already discriminated before this event. The government of the United States tried to ensure the Arabs were politically voiceless in the American civil society (Abraham, 1994). This was particularly the case during the second half of the twentieth century. Arab immigrants, who came to America in the beginning of the twentieth century on the contrary, were integrated more smoothly into American society. This was mainly because most of them were Christian, they were considered to be white and there was a minimal involvement of the United States in Arab countries (Cainkar, 2002). This changed later on, when the United States government tried to exclude them from the civil society. During the pan-‐Arabism, which was at its height during the 1950’s and 1960’s, Arabic people felt a very strong nationalistic feeling, and were proud to have an Arab identity even though this had some negative consequences on their position in American society. But after the Six-‐Day War in 1967, the pan-‐Arabism started to lose its impact and also the attitude towards the Arab identity changed (Ajami, 1978). Now some Arab immigrants or second-‐generation Arabs were changing their names from an Arab-‐sounding name to a more American-‐sounding name, from Farouq to Fred for example (Cainkar, 2002). Especially the college-‐graduates changed their names, probably because they suffered from the negative consequences of their Arab-‐sounding name. But being in between identities is not something a person desires. (Widner & Chicoine, 2011).

During the 1990s, things started to change: it was a global pattern. Arab people from all over the world started to identify themselves as Muslims, instead

(14)

of Arab. There was a shift from a secular to a religious identity. The secular political movements had failed so far, so they lost faith (Cainkar, 2002). They share a feeling of powerlessness as a minority and if Muslims of all different kind of minorities assemble as Muslims, they will feel less powerless. This has consequences for the level of discrimination.

After 9/11, it was not the minorities who were more hated, violated and discriminated, but it were Muslims. Therefore, because of the Black Power Movement, and thus influence of the Nation of Islam on distinctive black names, and the shift from secular to religious identity and people identifying themselves as Muslims instead of their origin, terroristic attacks organized by Islamic groups also increase discrimination of minorities such as black and Arabic people. This could be due to the generalization of an individual based on average information available of a particular group, which Arrow and Phelps describe as statistical discrimination. This is why black or Arabic people, with an Islamic-‐sounding name, may not be discriminated because they are black or Arabic, but because people assume they are Muslim. The cause of the discrimination in this case may be based on Becker’s prejudice-‐based discrimination model, but the increase of the discrimination after increase of terroristic attacks may be due to statistical discrimination. People might sooner associate black or Arabic people with being Muslim after more terroristic attacks by Muslims that draw more negative attention towards the Islam.

Racial structure changing over time

Historically, in the United States there exists a racial classification, which involves some form of hierarchy. Bonilla-‐Silva (1997) writes the race that is placed at the top is favourable on the labour market, has access to better occupations, and is thought to be smarter and better looking. All these social relations and practices, based on race, constitute the racial structure of society. There is difference in social, political and economic rewards to groups along racial lines, which is socially constructed. This does not mean every person within a particular race receives the exact same level of rewards and not all members of the race that is placed at the top of the society are actually at this social top. There is variance between the members of the race. However, races, as a social group, take a certain position in the social system as a whole (Bonilla-‐ Silva, 1997).

In the United States, white people have always been on top of society. Black people are at the bottom of society and Hispanics, Arabs and other groups are in between. However, the level of discrimination is changing over time. In the past century there is a decreasing trend in discrimination showing. For example, the term of people with a dark skin colour changed over time. It started with Negro, changed to Black and eventually to African American. This is perceived as less discriminating (Bonilla-‐Silva, 1997). In 1985, just 37 percent of the white people said they would ever vote for a black person if he were running for president and qualified enough. In 1997 96 percent stated they would do so (Williams, 1999) and from 2009 until 2017 the United States actually had a black

(15)

man1_{as
their
president.
Nevertheless,
there
still
is
a
strong
persistence
of} stereotyping, in a negative way. For example, related to the labour market at the end of the twentieth century, data show 29 percent of the white people think most black people are unintelligent, 45 percent think they are lazy, 51 percent believe most black people are more violent and 56 percent indicated most black people prefer living off welfare (Williams, 1999).

Socio-‐economic status

Data shows lots of minorities all live together in the same neighbourhoods. This has several causes and consequences. For example, from 1900 until 1940 only the least desirable neighbourhoods were available for African Americans, which caused a physical separation of blacks from whites (Williams, 1999). The dataset of Betrand & Mullainathan (2004) also shows different zip codes are still mainly white or mainly black, probably as an intact consequence of this previous structure of segregation. To take another example, Vietnamese immigrants came to the United States as a consequence of political oppression in the 1960s during the Vietnamese War. They clustered all together in the same neighbourhoods. This took away the need to integrate with the American culture or to learn to speak their language. Because of this, minorities are likely to hold on to their own culture and beliefs and therefore it is reasonable to find homogeneity within a minority in their socioeconomic status (Sharpe & Abdel-‐Ghany, 2006).

Minorities that are clustered together in the same neighbourhoods are more likely to have a lower socioeconomic status because overall they are less educated and have less access to well-‐paid employment. This contributes to the stereotyping of minorities and thus discrimination of an individual, prejudiced to be exactly the same as the average member of their minority as a social group. Asian Americans however, tried to change their socioeconomic status in society by overachieving in education. This resulted in Asian people to be closer related to the socioeconomic status of white people, but their wages are still lower compared to white people with the same level of education (Sharpe & Abdel-‐ Ghany, 2006). Also other early immigrant groups, such as Irish Catholics, Italians and Jews show a pattern in which they faced discrimination, but gradually attained an equal economic status to whites (Cain, 1986). In line with these developments are the Hispanics. Because they are mainly recent immigrants, their discrimination by lower earnings could be explained by their low education, level of English language and integration. If they will develop consistent with earlier immigrants as the ones named above, one may expect to see a decrease in level of discrimination (Cain, 1986; Bonilla-‐Silva, 1997; Borjas, 2010).

Names can also be a signal for the socioeconomic status of an individual. Before the 1960s, black and white people gave almost similar names to their children. The median name that was given to black girls was twice as likely to be given to a black girl than to a white girl. During the Black Power Movement there was a trend of black people adopting increasingly distinctive names and the

1_{Even
though
Americans
may
say
or
feel
Obama
is
the
first
black
president} (Harris, 2012), this is not true according to terminology. Because his father is from Kenya and his mother from Kansas, he is biracial.

Discrimination on the labour market : is labour market racism mainly the result of racist preferences or uncertainty reduction through statistical generalisation?