The effect of gender and ethnicity on the risk of over-education on the Dutch labour market

1

The Effect of Gender and Ethnicity on the Risk of

Over-education on the Dutch Labour Market

Rosalie Lilipaly 10205381 rosalie.lilipaly@hotmail.com Dr. A.M. Kanas Dr. B. Lancee Master thesis Submitted on 30June 2016

3

It is commonly known that labour market inequality persists on the Dutch labour market. Research suggests that labour market opportunities are influences by factors such as gender and ethnicity. A key aspect of labour market success is how well matched the job is to one’s education and skills. When an individual has a level of education in excess of that which is required for their particular job the person is regarded as over-educated. Over-education is prevalent in many western countries. This thesis studies the effect of gender and ethnicity on the risk of perceived over-education on the Dutch labour market. Data of the Longitudinal Internet Studies for the Social sciences panel (LISS) is used to investigate the possible correlation between over-education and gender and ethnicity. Moreover, the effect of both gender and ethnicity will be studied together as a double disadvantage. In addition, mechanisms that might influence the effect of over-education amongst women and immigrants will be tested. Previous studies were utilised or the theoretical background and focused all on western countries. Results obtained in this thesis reveal that gender has no significant effect on the perception of over-education whereas the main effect of being an immigrant turned out to be significant and positive for both men and women. The double disadvantage has no significant effect. Moreover, the positive effect of non-western immigrants seemed to be the reason why immigrants overall have a higher perception of over-education. Furthermore it is shown that part-time work has a significant effect on over-education but does not particularly cause women to be more educated. Additionally field of study is found to have no significant effect on over-education overall. Moreover, the interaction effect of gender and cohabitation does have a positive effect on perceived over-education and no significant effect is found for female respondents who have children. Lastly the conclusion and limitations and recommendations for future research will be discussed.

5

First of all, I would like to thank my supervisor Dr. Agnieszka Kanas for her help and feedback. Another thanks goes to Bram Lancee for agreeing to be my second

supervisor.

Furthermore I would like to thank my classmates for helping and supporting me during my studies. Especially Rafael needs to be mentioned since he helped me through some statistical challenges during my master thesis. In addition, I would also

like to thank the department of sociology for broadening my knowledge.

Last but not least, I would like to say thanks to my family and friends for having faith in me. I am particularly grateful for the support of my parents who facilitated my

7

TABLE OF CONTENTS

CHAPTER 1. INTRODUCTION ... 9

CHAPTER 2. THEORETICAL FRAMEWORK ... 13

SECTION 2.1AN INTRODUCTION TO OVER-EDUCATION ... 13

SECTION 2.2GENDER DIFFERENCES ON THE LABOUR MARKET ... 15

SECTION 2.3 ETHNIC DIFFERENCES ON THE LABOUR MARKET ... 17

SECTION 2.4 THE DOUBLE DISADVANTAGE ... 19

SECTION 2.5FIELD OF STUDY AS AN EXPLANATORY MECHANISM ... 20

CHAPTER 3. DATA & METHODS ... 23

SECTION 3.1 DATA ... 23

SECTION 3.2METHODS ... 24

SECTION 3.3MEASURING OVER-EDUCATION ... 26

SECTION 3.4 VARIABLES ... 28

CHAPTER 4. RESULTS ... 33

SECTION 4.1DESCRIPTIVE STATISTICS ... 33

SECTION 4.2CONSTRUCTION OF THE MODELS ... 36

SECTION 4.3 INTERPRETATION OF THE RESULTS ... 36

CHAPTER 5. DISCUSSION ... 41

CHAPTER 6. CONCLUSION ... 46

SECTION 6.1CONCLUSION ... 46

SECTION 6.2LIMITATIONS AND RECOMMENDATION FOR FUTURE STUDIES ... 46

REFERENCES ... 48

9

It is common knowledge that not everyone has equal access to the labour market. Work related inequality has been studied many times by economist and social scientist, in particularly inequality concerning certain groups. Research suggests that labour market opportunities are influences by factors such as class, gender, age and ethnicity. Two key aspects of labour market success are whether or not one finds a job, and how well matched the job is to one’s education and skills. In this study we focus on the second aspect. Evidence from European studies have shown that both women and immigrants workers regularly experience problems once they are participating on the labour market, such as unfavourable terms of employment, less access to promotion opportunities, relative lower wages or less access to higher pay or training (Bisin et al., 2011) (Friedberg and Rachel, 2000) (Robst, 2007). Labour market opportunities are vital to the integration of immigrants into their host countries and for the emancipation of women on the labour market. Numerous studies have explored some of the disadvantages women and immigrants experience on the labour market, such as unemployment gaps and pay gaps. Few studies have examined the extent of over-education amongst women and immigrants and which factors play a crucial role in their risk of over-education. Therefore this study will examine the role of ethnicity and gender on the risk of over-education.

Over-education is a fairly recent phenomenon. Western countries have gone through a process of tremendous expansion of education, especially of higher education in the past century (Schofer et al., 2005). This process is regarded as a positive development by many, it might even be necessary in order to meet an increasing demand for skills associated with the occurrence of the ‘knowledge society’. In addition, one could argue that better educated citizens have a positive effect on the society as a whole since there is a positive correlation between education and productivity (Barone and Ortiz, 2010). In contrast, some researchers found that educational expansion has its negative effects. Studies have shown that the rise in educational attainment exceeds the demand of highly-educated workers in the labour market (McGuiness, 2006). Hartog (2000) suggested that ‘The strong expansion of participation has outpaced the increase in the demanded levels of education’ (134). This view demonstrates that educational expansion

10

also comes with the risk of devaluation diplomas and consequently social degradation of tertiary graduates. Each occupation can be viewed as having a required level of education that is needed. However, there may be workers within any occupation with an obtained level of education greater than the required level of education. These workers are classified as being “over-educated”. “Over-education describes the extent to which an individual possesses a level of education in excess of that which is required for their particular job” (McGuinness, 2006).

Over-education has been widely discussed in the sociology and economics of labour literature, it dates back to Richard Freeman’s ‘Overeducated American’ (Freeman, 1976). Freeman (1976) argues that an oversupply of individuals educated on the university level in the US since the beginning of the 1970s had resulted in a decline in return to education. The existence of over-education could impose a serious waste of individual and societal resources. It is potentially costly for the individual, the economy and the firm (McGuiness, 2006). For instance, national welfare is possibly lower than would be the case if the skills of all workers were fully taken advantage of within the economy. Furthermore, tax revenues may be wasted on supplying individuals with non-productive education (Idem.). Morrow and Johnson (2002) illustrated that when employees perceive themselves as over-educated, they have more negative job attitudes and are less satisfied with their jobs. This might explain the evidence suggesting that over-education is associated with lower productivity. In addition, Verhaest & Omey (2006) found that over-educated individuals are also more likely to leave the organization. Overeducated workers are likely to earn a lower return on their investment as compared to similarly educated workers whose job match their level of education. A proportion of their educational investment is unproductive. Moreover, previously well-matched workers in the economy can be ‘bumped down’ in the labour market or even ‘bumped out’ of the labour market entirely, due to the over-educated workers who move into lower levels of occupations. Therefore over-education can effect both highly educated individuals as lower educated individuals in society. By moving to lower levels of occupations the average level of education within these occupations will rise. Previously competently educated workers might therefore need to shift to occupations beneath their level of obtained education. Moreover, over-education rate appears to be increasing (Vaisey, 2006). Developed countries are commonly more confronted with a rise of over-education

11

on the labour market. This means that there is a loss of education investments because the acquired skills and education of an individual are not used efficiently. Therefore it is interesting to study if there is overinvestment in education from a policy perspective. Especially in the context of Western nations where education is heavily subsidized, it is important to know whether education investments pay off. Determinants of over-education and knowledge of this extend are consequently key for policymakers and researchers on labour market stratification and integration.

This thesis examines if some individuals in the Dutch society might be more likely, and thus more vulnerable, to be over-educated based on their ethnic origin and gender. Therefore, the main research question this thesis addresses is:

“To what extent do gender and ethnicity have an effect on the risk of (perceived) Over-education in the Dutch labour market?”

In this thesis I will assess the combined effect of gender and origin on over-education. Firstly, the differences in overeducation on the Dutch labour market between men and women will be discussed. The position between men and women on the labour market have often been studied. Main examples are the gender wage gap and occupational sex segregation. Researchers have often explained why women earn less than man, reasons are given to explain the gender wage gap such as work-family reconciliation. Based on these findings it would be interesting to examine whether women are also more likely to have qualifications such as education and skills that exceed job requirements. They might have a disadvantage when it comes to acquiring a job on their obtained level of education due to gender differences on the labour market. Secondly, I will discuss the differences in overeducation between ethnic minorities/immigrants and native Dutch workers on the Dutch labour market. As stated before, labour market opportunities are critical to the integration of immigrants into host societies. Given their disadvantaged labour market position in many developed countries, there is reason to expect that immigrants are disproportionately affected by over-education. Thirdly, the effect of gender and origin will be combined. Non-Native Dutch women might experience a ‘double disadvantage’ of both the negative effect based on their gender and on their

12

ethnicity on their labour market success. The hypothesis of this double disadvantage will be tested with reference to the risk of over-education.

Furthermore, I would like to examine mechanisms for the presumed higher level of over-education amongst females and ethnic minorities/immigrants. Therefore I would like to take into account the influence of field of study on the risk of over-education. Individuals who studied certain fields are more prone to over-education (Dolton and Vignoles, 2000). Previous research argues that women are more likely to have obtained education in fields of studies where the risk of over-education is higher. Field of study can explain over-education for both higher educated and lower educated individuals. Vocational training also has different fields of study which might explain a higher level of over-education amongst certain groups of respondents. In addition, as stated before lower educated individuals can ben ‘bumped down’ in the labour market or even ‘bumped out’ of the labour market entirely, due to the over-educated workers who move into lower levels of occupations.

In conclusion, I would like to examine whether variation in labour market over-education in terms of the probability of over-over-education can be attributed to personal determinants of individuals, such as gender and origin. The purpose of this study is to shed light on over-education in the Dutch labour market as it affects men and women with different kinds of backgrounds.

13

Earlier research and insights about the topic of over-education on the labor market will be discussed here. The studies used in this chapter are focused on different countries such as the United States and the United Kingdom and might be a bit dated.

Section 2.1. An introduction to over-education theory

There is a compelling portion of literature on overeducation within the context of existing views of the labour market (McGuinnes, 2006). The theories that are considered to explain over-education are mainly side theories, in this thesis also the supply-side theories will be used. Nonetheless, it could be argued that over-education is also a result of demand-side factors (Ortiz & Kucel, 2008). For instance, individuals might be overeducated due to varying demand for graduates across different sectors. Therefore the sector size before graduates got their diploma may affect their odds of being over-educated (Idem.). In this thesis the demand-side factors will not be taken into account. The principle of the concept of overeducation lies in the first supply-side theory which is the Human Capital Theory (HCT) (Becker, 1980). It states that education and training increases individuals’ productivity (Becker, 1964). Thus, individuals must hold jobs in which their education and skills are applied. This theory implies that the skills of workers are fully utilized by employers (Ortiz & Kucel, 2008). Also, according to the Human Capital Theory, wages will always be equate to the individual worker’s marginal product, this marginal product is determined by the level of human capital that they have accumulated through either on-the-job training and/or formal education. Thus, there shouldn’t be any over-education in a market equilibrium, especially on the long run. The Occupational Mobility Theory (OMT) supports the HCT by suggesting that some over-educated workers possess lower abilities than others, their over-education is therefore a compensatory mechanism for their lack of skills (Idem.). This approach illustrates why young workers are frequently over-educated due to the lack of experience. To conclude, we might argue that overeducation suggests that the Human Capital Theory is not consistent when we look at the literature about the existence of over-education. When an employee is over-educated we can state that the production capacity of the individual is not reached. Yet, there might be other factors such as the

14

occupational mobility theory which may explain differences in educational match on the labour market.

Another theory that has gotten considerable attention within the overeducation literature is the Job Competition Model (JCM). This theory is based on Lester C. Thurow’s (1975) book ‘Generating Inequality’. The model describes a market where individuals compete for job opportunities based on their relative training costs, as opposed to competition based on the willingness to accept wages given their human capital (McGuinness, 2006). According to this model, job characteristics may be the only factor determining earnings. Thurow (1975) argues that the majority of workplace skills are acquired through training on the job instead of formal education. In this view the labour market is a place where training is provided and must be allocated to different workers (McGuinness, 2006). Candidates compete witch each other for a job on the basis of their relative training costs. When an individual has more education there is less training required which results in a better position in the job queue.

The HCT and the JCM are seen as two extremes, in the middle of the previous theories you can find the assignment theory (AT) (Sattinger, 1993). Sattinger (1993) attempted to reconcile the two theories. The assignment theory draws attention to both individuals’ desires and job characteristics. Workers may obtain jobs which require less education that they have acquired, yet their maximum income and/or utility may still be fulfilled. It implies that here is no reason to expect that wage rates are entirely related to obtain schooling or individual attributes as the Human Capital theory states. Also, we should not expect that wage rates will be entirely related to the nature of the job as the Job Competition Model proclaims. The assignment theory assumes that the productivity, such as income and job satisfaction, of a particular job match is actively influenced by the match between the graduate characteristics and the job requirements and (Teichler, 2009). Thus, like the job competition model, Sattinger (1993) assumes that there are limited jobs available in the labour market, this implies that compensation is independent of human capital of the individual. On the other hand he assumes that when individuals invest in human capital they are able to compete for the best job and wages. Over-education emerges because neither individual attributes are related to wages, nor the job competition (Idem.).

15

Section 2.2. Gender differences on the labour market

Substantial labour market differences exist between women and men. For instance, it is an established fact that women earn less than men (Glauber, 2012). Many theories are used to explain gender differences such as the gender wage gap between men and women. It would be interesting to investigate to what extend previous labour market research regarding gender can be used to explain a difference in risk of being over-educated on the labour market between men and women.

Gender inequality on the labour market is often explained in terms of occupational sex segregation. Jobs are segregated by sex (Reskin 1993). The principle here is that men and women tend to work in different occupations. In addition, working in different occupations can highly affect the returns of labour. Men predominate the most

prestigious professions, upper managements while women dominate professions such as nursing and teaching (Idem.). We can distinguish horizontal and vertical

occupational sex segregation. Vertical sex segregation focusses on the different

opportunities between men and women, it states that men have better possibilities on the labour market and are able to attain positions of power more easily. In contrast, women can be shut out from these higher positions. Horizontal segregation means that men and women tend to work in different jobs while having the same level of education. The tendency for women to work in systematically different occupations and industries than men is an important feature of the modern labour market.

Evidence from the US demonstrates that this contributes to a substantial gap between the earning of men and women (Blau and Kahn, 2007). Also, this gives women poor access to most influential positions in companies. Additionally, women tend to make different educational choices when it comes to field of study and specializations, this in turn might influence their risk of being over-educated. This will be discussed more extensively in the next section.

The following hypothesis is made based on the discussed literature:

 Hypothesis 1: Female workers have higher risk of being over-educated on the Dutch labor market than men with the same level of education.

Furthermore, women tend to have breaks during their career because they might get pregnant and/or have to take care of their family. Traditionally, it is the responsibility

16

of the male in the Netherlands to earn the money on which the family lives, and the female’s responsibility to take care of the family. Therefore, women might fall behind men when it comes to obtaining (on-the-job) human capital. It has been sown that having children negatively affects the labour market outcomes, such as hourly wages, of women (Glauber, 2012). Men are able to capture more human capital during their careers because they are not out of the labour market for a longer period of time due to pregnancy or taking care of the family. In contrast, women tend to leave the labour market due to family reconciliations. Research has found that women accept jobs for non-market reasons such as household duties or illness in the family (Sicherman, 1991). Groot (1993) finds that over-educated workers have less experience, tenure and on-the-job training. Therefore over-education can be explained through a

compensation for a lack of other human capital endowments, such as on- the-job training or experience (Idem). Workers who experience a career interruption, such as women with children or household duties, are therefore more likely to be in jobs for which they are over-educated. Having children means that mothers in the Netherlands still have to do most of the caring tasks at home and therefore leave the labour market more than men. Frank (1978) links the higher probability of women being

overeducated to their limited geographical mobility in the US. Women tend to be the second earner in the household and their decisions are therefore subordinated to those of men, or their partner. Married women or women living with a partner may face more geographic constraints in searching for a job (Frank, 1978). As such, it would be expected that married women in small labour markets are more likely to be overqualified than men. Family mobility is a joint decision in which the needs of male and female are balanced to maximize family welfare (Robst, 2007). In addition, job-motivated relocations are generally made to benefit the primary earner in the household which lead to a constrained job search for the second earner since the second earner must search for a job in a limited geographic area. Traditionally the male is still the primary wage earner in households whereby often the women may be at disadvantage (Idem.).

17

The following hypothesis is made based on the discussed literature:

 Hypothesis 2: Female workers have particularly a higher risk of perceived over-education on the Dutch labour market when they have children and/or live together with a partner.

In addition, in the Netherlands a lot of women tend to work part-time. The labour force participation of women in the Netherlands was amongst the lowed in the OECD in the early 1980s (OECD, 2004). In the following decades the labour force participation increased substantially and this was due to the fact that more women were working part-time. Female part-time work has remained popular in the Netherlands. Even though this has increased the low female participation rates, it is also considered to have a negative effect on the potential of women on the Dutch labour market. Part-time working women are paid less and have fewer opportunities for promotion

according to research on the US labour market (Blau & Kahn, 2007). This might also effect the presumed higher risk of over-education for females on the Dutch labour market. When women don’t use their full potential on the labour market, they might also have a higher risk of being in a job for which they are over-educated.

Lastly, employers might discriminate women based on real or imagined statistical distinctions between male and female workers. Statistical discrimination occurs, for instance, when female workers are offered lower wages because women are perceived to be less productive on average compared to male workers. It might be the case that women are, on average, more over-educated on the Dutch labor market due to

discrimination. Unfortunately this cannot be tested in this research. The following hypothesis is made based on the discussed literature:

 Hypothesis 3: Female workers have particularly a higher risk of perceived over-education on the Dutch labour market when they work Part-time.

Section 2.3. Ethnic differences on the labour market

It is well known that ethnic minorities in Europe generally hold less favourable positions in the labour market than the majority of the population. There are ethnic gaps in employment and wage in western European countries (Heath et al, 2008). A factor that can contribute to foreign-native gaps in labour market outcomes are

18

disparities in human capital. Because ethnic minorities on average have relatively low levels of education, are less proficient in the host-country language and have less knowledge of labour market institutions partially explains their unfavourable position. Also cultural or social competence are used to explain labour market gaps based on origin. Social capital states that individuals can more easily access resources through network relationships. According to social capital theory (Coleman, 1988) individuals can have advantages based on their social resources. When someone originates or has parents who originate from another country, they might experience a loss of

community social capital. Furthermore, it is argued by several authors that the education that immigrants acquired in their countries of origin are less valued than skills obtained in the host country (Friedberg, 2000). The education or skills can be difficult to transfer or are of lower quality. Also, employers have more uncertainty about the quality of the education and skills. When immigrants acquire host-country human capital they may improve their economic position. Second generation

immigrants may therefore have another advantage over first generation immigrants. The dataset classifies immigrants into western and non-western immigrants. In the Netherlands the largest non-western ethnic minority groups are Moroccans and Turks (Blommaert et al., 2014). These groups differ with western ethnic minorities with respect to their social-cultural integration. Western immigrants might experience several advantages such as cultural and societal competences because it is more difficult to integrate socially and culturally for non-western immigrants. Hence, the western immigrants might have an advantage over non-western immigrants when it comes to overeducation risk (idem).

In addition, there are two types of discrimination which might affect the prospects of individuals based on their origin. Hiring practices can be shaped due to statistical discrimination and taste discrimination. The theory about taste discrimination suggests that employers are willing to pay more for the labour of members of a

preferred group. The explanation is that they might experience social or psychological distance from other groups. Furthermore, statistical discrimination is described as a choice by the employer not to hire members of a certain group or origin based on the belief that this groups is, on average, less productive than other groups. When a job

19

applicant has a foreign-sounding name he or she might not be hired by a company because the recruiters assume or are uncertain about the quality of the language skills of the individual. However, it is not possible to control for discrimination in this research due to data availability constraints.

The following hypothesises are made based on the discussed literature:

 Hypothesis 4: Workers with non-Native Dutch origin have higher risk of over-education on the Dutch labour market than workers with a Dutch origin with the same level of education.

 Hypothesis 5: Workers with a non-western ethnic origin have higher risk of over-education on the Dutch labour market than workers with a Dutch or other western ethnic origin with the same level of education.

Section 2.4. The Double Disadvantage

As stated earlier, the literature on international migration contends that immigrants often experience substantial hardships when entering the labour market (Raijman & Semyonov, 2997). Two main difficulties are the ability to join the labour market and the ability to find a rewarding and suitable job (Idem.). However, there seems to be a difference between gender lines when it comes to the size of the presumed advantages on the labour market. Boyd (1984) was the first who articulated the double disadvantage to describe the labour market position of immigrant women, specifically in Canada. According to Boyd (1984) immigrant women’s labour force position reflects the combined negative impact of sex and birthplace. They are therefore doubly disadvantaged in comparison with native born females and men. She found that even though immigrant women participated in the labour force more than Native-Canadian women, immigrant women were disadvantaged in terms of occupational prestige. Boyd gave the following definition of the “double disadvantage’:

“Sex adds another dimension to the stratification of immigrants within the workplace and within the larger society. In addition to the status of being a migrant, immigrant women experience additional difficulties in the labor force as women... Overall, the position of

20

immigrant women in the labor force can be understood as reflecting the combined impact of sex and birthplace or the "double negative" effect.” (1984, 1092-3)

The double disadvantage will be tested in this thesis with regard to over-education. It might be the case that female ethnic minorities are more prone to over-education due to their double disadvantage. The group of female immigrants is compared to all immigrants and to all females in the receiving country. In addition, female immigrants from less developed countries earned less than immigrant females from more developed nation in the 1984 study (Boyd, 1984). You might therefore argue that immigrant from poor nations experienced more strongly negative effects related to the double disadvantage. In this thesis will test if women with a non-Western origin might experience higher levels of over-education as compared to women with a western-origin. Other authors have also testes the double disadvantage hypothesis and found for instance that immigrant women with a Hispanic origin had a disadvantage in the United States when compared with Hispanic immigrant men and native women (Raijman & Semyonov, 2997). Therefore they were less successful in converting their human resources, such as education, into occupational prestige. Moreover, Katz (1982) argued that female Russian immigrants suffered from this double disadvantage when competing for jobs demanding high levels of education in Israel.

 Hypothesis 6: Female workers with a non-Native Dutch origin have higher risk of over-education on the Dutch labour market than workers with a native Dutch origin.

Section 2.5. Field of study as an explanatory mechanism

Traditionally privileges and labour market opportunities were associated with for instance the attainment of a University degree. Nonetheless, it seems that this is no longer generalized but limited to a selection of well-established institutions and study programmes (Berggren, 2008). Depending on the length and field of a study the programme comes with prestige (Capsada-Munsech, 2015). The definition of prestige regularly used is that of Davies and Gruppy (1997), it is based on the level of expected economic returns after graduating. They have found that the field of study ‘engineering’’ comes with higher economic returns, on the other hand home economics had the least

21

economic returns. Furthermore, economic and business are above median and social sciences and humanities below. Whereas physical and earth sciences are in the median. Moreover, it has been established that the best-performing fields in the past two decades remained virtually the same according to a study in Italy (Ballarino & Bratti, 2009). Considering these findings I believe that taking into account fields of study into the analysis of over-education may give us more insight. Field of study might serve as an important mechanism regarding over-education.

Ortiz & Kucel (2008) state that fields of study can be understood as different stocks of human capital, differentially valued by companies. It is possible that employers don’t value human capital stock represented by one field as much as the human capital stock represented by another field. Therefore the less valued field of study might suffer from a higher incidence of over-education. Contrasting employees with fields of studies which are well recognized by companies. These employees might therefore experience over-education less frequently. Employers have several reasons why they may value the human capital obtained by an individual’s field of study differently (Idem.). One reasons is that the average duration of studies might vary within a given field and level, both theoretical as in years. Another reason is that fields of studies intrinsically require a degree of specialization, however some fields are more occupationally focused than others (Ballarino & Bratti, 2009). For instance degrees within ‘Health’ are evidently more aimed at certain employments, whereas other fields as ‘Arts’ are less specific, this leads to a wider range of employments (idem.). Therefore over-education might differ across fields of study. In addition, Dolton and Vignoles (2000) have studied graduates in the UK and found that a huge portion of graduates are over-educated after graduation. Supplementary, they also found that individuals who studied certain fields were more prone to over-education. These academic programs include social sciences, arts and languages. In contrast, graduates who studies engineering or technical studies were less prone to be over-educated. To conclude, while education always served as a distinctive factor for privileged individuals to keep their high position on the labour market, nowadays individuals are more in need to find themselves some qualitative distinctions within particular fields of education in order to obtain a privileged position. Fields of study might therefore serve as an important distinctive factor when studying over-education. The expansion of education and its impact on the labour market may have

22

increased the relevance of field of study when exploring over-education.

Women and men pursue different fields of study in college. Women tend to choose studies such as social sciences more frequently, which can result in a higher risk of over-education. As stated before some fields less specific, this leads to a wider range of employments (Ballarino & Bratti, 2009). Previous research has found that these less specific field give individuals a higher risk of being over-educated (Idem.). In contrast women are underrepresented in science and engineering (Green, 1989). Different choices in field of study between males and females might have important economic and social impacts (Zafar, 2008). Therefore it would be interesting to study the influ-ence of field of study on the presumed risk of over-education amongst women in the Dutch labour market. Field of study is especially relevant in studying highly educated female workers. Gender differences in degree may arise for several reasons, such as differences in types of subjects male and female students study. These gender differ-ences in attainment might arise because of psychological and/or biological factors. Furthermore, these differences might be the result of male and female stereotyping or prejudice.

 Hypothesis 7: Female workers have particularly a higher risk of perceived over-education on the Dutch labour market because of their field of study.

Moreover, recently the CBS (Het Centraal Bureau voor de Statistiek, 2016) reported that highly educated unemployed individuals with a Dutch origin find work more easily as compared to high educated unemployed individuals with a non-Dutch origin. In addition they stated that this cannot be explained by field of study. Individuals with a non-Dutch origin tend to study classical studies, such as technical studies, which gives them favourable job opportunities. In contrast, according to the CBS (2016), individuals with a native Dutch origin are more likely to do humanitarian studies such as psychology and languages. These studies have less favourable job opportunities. Thus it is interesting to examine whether field of study also does not have an effect on the overeducation risk based on origin.

 Hypothesis 8: Workers with a non-Dutch origin on the Dutch labor market don’t particularly have higher risk of perceived over-education on the Dutch labour market because of their field of study.

23

Chapter 3. Data & Methods

Section 3.1 Data

For the research conducted in this thesis the data of the LISS (Longitudinal Internet Studies for the Social sciences) panel administered by CentERdata (Tilburg University, The Netherlands). The research is based on secondary data (Bryman, 2012). Second-ary analysis offers several benefits such as having access to high-quality data, and saving costs and time. Yet, limitations of using secondary analysis are for instance the lack of familiarity with the data and the complexity of the data. Sometimes, the sheer volume of data can cause problems with the management of information (ibid.). In this study the LISS data is used because there is the opportunity to study quite sizable subgroups. This data set yields quite large representative samples. The LISS panel is a representative sample of Dutch individuals who participate in monthly Internet sur-veys. The panel is based on a true probability sample of households drawn from the population register. Households that could not otherwise participate are provided with a computer and Internet connection. A longitudinal survey is fielded in the panel every year, covering a large variety of domains including work, education, income, housing, time use, political views, values and personality. The LISS panel is the principal com-ponent of the MESS project The LISS panel is intended for scientific, policy or socially relevant research. The quality and the coverage of the sample is of prime concern. The original LISS panel exists since October 2007. Respondents receive monthly questionnaires on various topics. The respondents are also required to fill out the 'background' questionnaire. Hereafter respondents receive invitations to several

questionnaires that complement the data from the 'background' questionnaire. In total the original LISS panel consists of 5,000 households. The LISS immigrant panel was established in October 2010 and has been withdrawn since December 2014, unlike the still ongoing LISS panel. The immigrant panel also used a 'background'

questionnaire which includes questions on background information of the respondent. This background questionnaire was substantially identical to that of the normal LISS panel. Respondents received, as in the original LISS panel, invitations different

questionnaires. In total, the immigrant panel consisted of approximately 1,600 respondents, including 818 non-Western and Western 799 immigrants.

24

In this study I will utilize the two databases: The LISS panel (Longitudinal Internet Studies for the Social Sciences) and the Immigrant panel including the background variables of both datasets. The chosen dataset is about Work and Schooling. The questionnaire is part of the LISS Core Study, a longitudinal survey delivering a broad range of social core information about the panel members. The survey focuses on labour market participation, job characteristics, pensions, schooling and courses. The background variables need to be added separately into SPSS. I’ve combined the LISS data panel “work and schooling” wave 7 with the Immigrant panel “work and

schooling” wave 2. These waves were both collected in 2014. Wave 2 of the immigrant panel is the most recent wave available, therefore this wave is used together with wave 7 of the LISS data panel. In addition, I’ve added the background variables of the LISS data panel and the Immigrant panel. The background information of the respondents is used of the month before the core modules were conducted. When the two surveys are combined there is a total sample of approximately 10,000 respondents. However, obviously there are non-responses within the dataset and due to the design of the questionnaire not everyone did respond to all the questions. When a respondent is too young or too old to have an occupation, occupational questions are left out in that particular survey. As a result the number respondents does not add up to all the, but to approximately 4,000.

The data sample used in this research consists of individuals who are employed are between 25 and 65 years old. Also, the sample should only include individuals whose primary occupation is paid work. The dependent variable in this research are meas-urements of perceived over-education. The independent variables in this research are gender, origin and field of study. The control variables that are being used in this study are highest level of obtained education, age, and (semi-)public sector. Section 3.2 Methods

To answer the research question, the computer program SPSS is being used to do a statistical analyzes. By means of a logistic regression analysis, the effect of gender and origin on over-education can be measured. By adding a various types of variables that might explain the variance, we can see which factors are relevant mechanisms when

25

measuring the effect of gender and origin on over-education risk. Regression methods are widely used for analyzing the relationship between a dependent variable and one or more independent variables. The most popular regression method is linear regres-sion. It is however applicable if the dependent variable is continuous, independent and identically distributed only. In cases where the dependent variable is categorical, con-ventional regression analysis is not appropriate.

A binomial logistic regression, often referred to simply as logistic regression, predicts the probability that an observation falls into one of two categories of a dichotomous dependent variable based on one or more independent variables. The dependent varia-ble (y) is a dummy variavaria-ble (coded 0, 1), commonly the generic terms success and fail-ure are used for these two outcomes (Agresti & Finlay, 2009). The goal is to under-stand the relationship between the explanatory X-variables and the dependent Y-vari-able. Logistic regression generates the coefficients (and its standard errors and signifi-cance levels) of a formula to predict a logit transformation of the probability of presence of the characteristic of interest:

A proportion (p) refers to the fraction of the total that possesses a certain attribute. For instance, suppose we have a sample of four people. Two of them have indicated that they perceive to be over-educated. Therefore, the proportion of perceived over-educa-tion is 2/4 or 0.50. The logit transformaover-educa-tion is defined as the logged odds. The odds of success are defined to be:

If the probability of a presence of a characteristic = 0.75, then the probability of ab-sence equals 1 – 0.75 = 0.25, and the odds of success = 0.75/0.25 = 3.0. Whereas if P = 0.25, then odds = 0.25/0.75 = 1/3. The odds are nonnegative, with value greater than 1.0 when a success is more likely than a failure (Agresti & Finlay, 2009).

26

When a logistic regression is calculated, the regression coefficient (b1) is the estimated increase in the log odds of the outcome per unit increase in the value of the exposure. In other words, the exponential function of the regression coefficient (eb1_{) is the odds}

ratio associated with a one-unit increase in the exposure. Odd ratios are used to com-pare the relative odds of the occurrence of the outcome of interest, given exposure to the variable of interest. It can also be used to determine whether a particular variable is a risk factor for a particular outcome. The magnitude of various risk factors can be compared: OR=1 exposure does not affect odds of outcome, OR>1 exposure associated with higher odds of outcome, OR<1 exposure associated with lower odds of outcome. The odds ratio has a minimum value of zero but have no upper limit. A value less than one indicate that the case is not likely to prevail under those circumstances and a value greater than one indicates a high likelihood to prevail under those circum-stances. The further the odds ratio is from one, the stronger the relationship. The outcome of the regression is not a prediction of Y value, as in linear regression, but a probability of belonging to one of two conditions of. A further mathematical transformation is needed to normalise the distribution, this is called the log transfor-mation.

Section 3.3. Measuring over-education

Over-education can be distinguished by classifying it into ‘objective’ or ‘subjective’ defi-nitions. There are four primary approaches to measuring required education for a job and thus over-education, two objective measures and two subjective measures. The choice of measurement is generally restricted by data availability.

Overeducation can be assessed subjectively by asking the respondent to give

Information on the minimum requirements of the job and then comparing this with the individual’s acquired education or by simply asking the respondent whether or not they are overeducated. Overeducation can also be assessed objectively by using infor-mation provided by professional job analysts (such as in the Standard Occupational Classification System in the UK or the Dictionary of Occupational Titles in the US) to determine an individuals required education on the basis of their job title and again comparing this with their actual level of education. A second objective measure of

27

overeducation is obtainable by calculating the mean education level for a range of oc-cupations with an individual defined as being overeducated if they were more than one standard deviation above their occupation’s mean education level. Overeducation can be assessed subjectively by asking the respondent to give information on the minimum requirements of the job and then comparing this with the individual’s acquired educa-tion or by simply asking the respondent whether or not they are overeducated. Overed-ucation can also be assessed objectively by using information provided by professional job analysts (such as in the Standard Occupational Classification System in the UK or the Dictionary of Occupational Titles in the US) to determine an individuals required education on the basis of their job title and again comparing this with their actual level of education. A second objective measure of overeducation is obtainable by calcu-lating the mean education level for a range of occupations with an individual defined as being overeducated if they were more than one standard deviation above their occu-pation’s mean education level.

A subjective mismatch is based on self-reports by individual workers. It is typically done by asking the respondent whether or not they are over-educated. This is called a direct self-assessment (DSA). Another subjective measure is to ask the respondent to give information about the requirements of the job, sequential the job requirements can be compared to the individuals’ obtained education (McGuinness, 2006). This is called the indirect self-assessment (ISA). The objective assessment of over-education can be done by using information provided by professional job analysts. By comparing years of education attainment with the average education level within the occupation of the worker (McGuinness, 2006). This method is called the normative/job analysis (JA) method. The second objective measurement of over-education is the statistical/re-alised matches (RM method and is retrievable by calculating the mean educational level for different occupations with an individual defined as being over-educated. A re-spondent is over-educated when the individual is more than one standard deviation above their occupation’s mean education level (Idem.).

Both subjective and objective measures have been criticized on a number of grounds. Subjective measurements are for instance criticized because individuals who work in smaller organizations may not have an inadequate benchmark against which to assess

28

their job requirements. This can produce a measurement error (Green et al., 1999). Secondly, the critique is that respondents may apply different criteria when assessing over-education Thirdly, over-educated workers may be less likely to respond to ques-tionnaires due to higher level of job apathy.3 Therefore the results may be an underes-timation of the incidence of over-education.. Nonetheless, Green et al. (1999) found in the UK that in most cases the respondents of education levels one needed to do the job tended to match the education levels to get the job. These results show that there is a high level of correlation in these subjective approaches. Disadvantages of the objective measurements are for instance that occupations may contain a number of skill levels, therefore people with the same job titles may be doing very different jobs (McGuinness, 2006). For example, responsibilities of managers are expected to vary widely. Further-more, if a particular occupation contains a high proportion of overeducated workers, the average occupational average will raise. Thereby the true level of over-education is overestimated. There might also be a high change of measurement errors when trans-lating job requirements to a single school variable when doing an objective measure-ment, also the method has an inability to capture the dynamic nature of job structure (Hartog, 2000). Very little new information is added. Previous literature has treated the concept of over-education as a negative phenomenon, also on the individual level. Morrow and Johnson (2002) illustrated that when employees perceive themselves as over-educated, they have more negative job attitudes and are less satisfied with their jobs. Verhaest & Omey (2006) found that individuals who perceive themselves as over-educated are also more likely to leave the organization. Because the perception of the individual is of high importance for these negative consequences it might be important to measure over-education subjectively. When individuals perceive that they have a mismatch it might highly effect the behaviour of an individual. In this thesis the per-ceptions of individuals are used when measuring mismatches on the labour market. The choice of this measurement is also limited by the data availability.

Section 3.4 Variables

As already discussed earlier in this chapter, the data of the LISS (Longitudinal Internet Studies for the Social sciences) panel administered by CentERdata (Tilburg University, The Netherlands) will be used. Firstly, there has been made a selection of the data by deleting the unemployed and those that do not want to work. These respondents are

29

not relevant and can’t be used for this analysis. Also, the respondents were selected based on their age, the following age group is being used in the analysis: 25 ≥ of ≤ 65. We restrict the analyses to respondents older than twenty-four of age but not older than sixty-five years of age because many respondents younger of age than twenty-four are still in education whereas many respondents older than sixty-five years of age have left the labour market for one reason or another. Of the original dataset 2.353 re-spondents remained when I applied a filter to the data set based on age and employ-ment. The main variables used in the analysis are defined below. Secondary analysis entails the analysis of data collected by others, therefore some key variables have to be transformed into different variables so we are able to test hypothesizes.

The dependent variable in this research is perceived over-education (also referred to as over-education). To measure over-education we use variable a variable where respond-ent can subjectively answer if their education matches their job requiremrespond-ents. Re-spondents are able to give five different answers when it comes to their level of educa-tion and match with their job. The statements that can be made in this measurement are ‘my education is’: (I) is approximately at the level required by my work, (II) is higher than the level required by my work, (III) is lower than the level required by my work, (IV) is for another kind of work than my current work, (V) has become outdated because the work has changed, (VI) has no relation at all to my current work, (VII) is insufficiently geared to the work practice, (VIII) I don’t know. To make this variable suitable for the analysis in this research it is transformed into a different dummy vari-able where 1 is ‘higher than the level required by my work’ and 0 is (I), (III), (IV), (V), (VI) and (VII). The last category (VIII) is due to the small number coded as system missing. Important is to note that over-education is not only opposed from perfect match, but also from other forms of ‘mismatch’.

The independent variables in this research are most importantly gender and origin. Field of study is also taken into account as an independent variable, as explained in section 2.5 this can be an important explanatory mechanism. Furthermore, the varia-bles ‘children’, ‘cohabitation’, and ‘part-time work’ are used in this research. The gen-der variable originally had a 1 and 2 structure, this is changed into a dummy, a 0 for male and 1 for female.

30

Origin is categorized in two different values: (I) Dutch background, (II) immigrants. CBS uses as a definition of immigrant called ‘Allochtoon’ (in the Netherlands). This is a person who lives in the Netherlands, of whom at least one parent was born abroad (CBS, 2016). A distinction is made between people who are born abroad (first genera-tion) and persons born in the Netherlands (the second generagenera-tion). The CBS also makes a distinction between Western and non-Western immigrants. Immigrant whose ethnic background is one of the countries in Europe (excluding Turkey), North Amer-ica and Oceania, or Indonesia or Japan are referred to as western immigrants. Based on their socio-economic and socio-cultural position, immigrants from Indonesia and Japan are among the westerners. It is mainly about people who were born in the for-mer Dutch East Indies and employees of Japanese companies and their families. Im-migrant whose ethnic background is one of the countries in Africa, Latin America and Asia (excluding Indonesia and Japan) or Turkey are referred to as non-western immi-grants (idem.). As already discussed in section 2.4, the western immiimmi-grants might have an advantage over non-western immigrants. Therefore this distinction is also used in this research. To make the variable ‘origin’ suitable for the analysis in this re-search it is transformed into a different variable called, where a distinction is made be-tween western immigrant (I), non-western immigrant (II) and native Dutch origin (III). The definition of ‘allochtoon’ by the CBS is referred to as ‘immigrants’ in the thesis. Furthermore three different dummy variables where created from this variable, a west-ern immigrant dummy variable, a non-westwest-ern immigrant Dummy variable and a na-tive Dutch dummy variable.

To measure field of study the variables answering the following question have been used: ‘in what field did you complete your highest level of education’. The original vari-able included 18 different categories. I recoded these into six categories: (I) general or other, (II) social sciences (including law & economics), (III) professional education, (IV) Humanities & Arts (V) Sciences and (VI) more than one fields. The answer ‘I don’t know’ is not included, it is reported missing due to small number of respondent an-swering the question with this variable. The following table presents the variables sorted in the six different categories.

31

To measure whether a person works part-time or full-time, the variable answering the following question has been used: ‘How many hours per week are you employed in your job, according to your employment contract?’ This variable is made into a dummy variable where part-time work is 1 and full-time work is 0. In addition, part-time work is regarded as working less than 36 hours a week, whereas full-time work is regarded as working 36 hours or more in a week. Having children had a yes or no setup and was changed into a dummy variable where 0 means that the respondent has no chil-dren and 1 means that the respondent has one or more chilchil-dren. Furthermore, the variable ‘Domestic situation’, has been recoded into the dummy variable ‘cohabitation’. Single and Single, with child(ren) are coded as 0, whereas (un)married co-habitation, without child(ren) and (un)married co-habitation, with child(ren) are coded as 1. The control variables used were highest level of education, age, and (semi-) public sec-tor. The original variable of highest level of education with diploma included 9 catego-ries ranging from, primary school to having a WO degree. I recoded these into five cat-egories: (I) primary education, (II) lower secondary education (Dutch: VMBO), (III)

Category Variables

I general or other general or no specific field, other area II social sciences

(including law & economics)

social and behavioral studies (including organization studies, media, culture, sports and leisure studies, etc.), economics, management, busi-ness administration, accountancy, law, public administration

III professional education

aagriculture, forestry, environment, medical health services, nursing, etc., personal care services, catering, recreation, transport, logistics, tel-ecommunication, public order and safety (police, army, fire brigade, etc.) IV Humanities &

Arts

Teacher training or education, art, humanities (modern or classical lan-guages, history, theology, etc.)

V Sciences mathematics, physics, IT, technology

VI more than one fields

32

higher secondary or vocational education (Dutch: havo/vwo or MBO), (IV) higher pro-fessional education (Dutch: HBO) and (V) University (Dutch: WO). Moreover, we con-trol whether the respondents have followed a training the las 12 months, variable cw14g035. The following question is asked: ‘Have you, in the past 12 months, followed any educational programs or courses or are you presently following one or more edu-cational programs or courses?’ This concerns eduedu-cational programs or courses that are important for the respondents work or profession. The variable originally had a 1 and 2 structure, and is recoded into a dummy variable, a 0 for ‘no’ and for ‘yes’. Fi-nally, the sector in which a respondent works in is controlled for. The variable meas-uring the type of organization is being used as a control variable. The variable also originally had a 1 and 2 structure, this is changed into a dummy variable, a 0 for pri-vate company and 1 (semi-) public sector.

33

Section 4.1 Descriptive statistics

The descriptive statistics can be found in table 1 below. The purpose of descriptive statics is to summarize data, to make it easier to assimilate the information (Agresti & Finlay, 2009). The total number of respondents is 2,353. It is easily seen that the majority of the sample consists of respondents with a Dutch origin, as many as 944 males and 1096 females. In addition, the sample of non-native Dutch respondents is rather small, with only 174 males and 136 females.

When looking at the dependent variable, over-education, we can already notice that the average percentage of perceived over-education shows almost no difference when looking native Dutch males and females. Only 19% of males with a native Dutch origin perceive themselves as being over-educated, compared to only 18% of native Dutch females. Furthermore, when we look at male respondents we can state that western immigrants slightly perceive themselves more over-educated as compared to native Dutch males. In addition, non-Western male immigrants have the highest percentages of perceived education, 30% of these respondent perceive themselves as over-educated. Additionally when looking at the female respondents we can see that there is also little difference in perceived over-education when comparing female native Dutch respondents and female western immigrants. Moreover, the same amount of western female and male immigrants perceive themselves as overeducated,

respectively 22%. Remarkable is that only 23% of female non-western immigrants perceive themselves as over-educated whereas 30% of male non-western immigrants perceive themselves as over-educated. In contradiction with the hypothesis it shows that non-western male immigrants are relatively more over-educated as compared to female non-western immigrants.

The independent variables in the descriptive table shows that most of the Dutch native females are working part-time, as many as 77%, whereas only 15% of Dutch native males are working part-time. Noteworthy is that non-native Dutch females are relatively working less in part-time jobs as compared to the native Dutch females. From all the female western immigrants 58% is working part-time and from all the

34

female non-western immigrants only 40% is working part-time. It is very clear that females overall are working more in part-time jobs. The second independent variable shows that either males and females mostly have children, all report a percentage above 50%. Native Dutch males and females show the highest percentages. Moreover we can see that native Dutch male and females have a particularly high percentage of cohabitation, respectively 81- and 75%. Male immigrants also show a high rate of cohabitation, 74% amongst western immigrants and 69% amongst non-western immigrants. The immigrant females are distinguished by lower rates of cohabitation, 55% amongst western immigrant women and 54% amongst non-western immigrant women. Other independent variables are the field of study. Especially interesting for this research is the high average of males in the field of sciences as compared to females, each category shows a percentage higher than 30%. Non-western male

immigrants in particular have a very high percentage in sciences, as many as 41%. In contrast, females are relatively underrepresented in the study of sciences, each female category shows a percentage lower than 9%. Furthermore it is notable that 46% of non-western immigrant females have chosen a study in social sciences. Moreover, overall more females chose a study in humanities and arts as compared to males. These results are in line with our theoretical framework.

Lastly, the control variables show, among other variables, the highest level of attained education. Since we include primary school in the highest level of attained education, the variable field of study might cause a problem considering that primary school has no fields of study. Nevertheless, primary school includes a very small amount of respondents and therefore this is not an issue.

36

In this section the results of ten binary logistic regression models are presented. The first model basically only provides the control variables, followed by model 2 which adds gender. The following models add important independent variables and some of them add important independent variables and its interaction variable(s). Model 1: Control variables

Model 2: Gender + control variables Model 3: immigrants + control variables

Model 4: Western immigrants + non-western immigrants + control variables. Model 5: Gender + part-time work + part-time work*gender + control variables Model 6: Gender + children + gender*children + cohabitation + gender*

cohabita-tion + control variables

Model 7: Field of study + control variables

Model 8: Gender + field of study + gender*field of study + control variables

Model 9: Gender + western immigrants + non-western immigrants + gender*west-ern immigrants + gender*non-westgender*west-ern immigrants + control variables Model 10: immigrants + field of study + immigrants*field of study + control

varia-bles

Section 4.3 Interpretation of the results

The logistic regression results can be found in table 2.1 and 2.2. Table 2.1 provides the result of models 1 through 5 and table 2.2 provides the results of models 6 through 10. The analysis was conducted to predict the risk of perceived over-educa-tion. Every model shows a significant χ2 _{which means that every model tested in this}

analysis is a better fit as compared to the null model.

The first model clearly shows that level of obtained education has no effect on per-ceived over-education. Furthermore, for every year a respondent’s age increases, the odds of perceived over-education decreases with 1.2%. Based on these results you could state that in general older respondents are less likely to perceive themselves as over-educated as compared to younger respondents. In addition, it is predicted that

37

the odds of perceived over-education decreases 25.4% when the respondent is working in the (semi-) public sector.

Furthermore we can see that the second model indicates that gender has nog signifi-cant effect on over-education. Consequently, hypothesis 1 is rejected. Despite the re-jection of hypothesis 1, the second and third hypothesizes were tested. Model 5 illus-trates that it is predicted that the odds of perceived over-education increases with 53.5% when the respondent is working part-time. Despite this result there is no signif-icant difference found when looking at the interaction effect between gender and part-time work. As a consequence, hypothesis 2 is rejected. Moreover, adding these varia-bles slightly increases the negative coefficient and significance of the variavaria-bles age and (semi-) public sector. The reason behind the latter effect is unclear. In addition model 6 adds the children and cohabitation variable to gender and the control variables. We can say that the odds of perceived over-education of a respondent who has children decreases with 27.9% as compared to respondents who don’t have children. Again the interaction effect, in this model concerning gender and children, has no significant ef-fect. When looking at cohabitation it is predicted that the odds of perceived over-edu-cation decreases 43.2% (with .99 level of confidence).Therefore cohabitation has a neg-ative effect on perceived-overeducation. In contrast, the interaction effect of gender and cohabitation shows a positive effect. Women who are living with a partner are 2.175 times more likely to perceive themselves as over-educated, this can be state with a 0.99 level of confidence. Consequently hypothesis 3 is partly true, when looking at cohabitation the odds of perceived over-educated increases for women. Yet there is no significant effect when looking at the interaction effect of female respondents who have children. Furthermore it is notable that model 6 doesn’t show a significant result for the age control variable.

The third model illustrates the logistic effect of the dummy variable immigrants on over-education. The model shows again an exp(B) value indicating that when age is raised by one unit, the odds ratio decreases with 1,2% and therefore less likely to be over-educated. Furthermore the model four shows that respondents with a non-native Dutch origin are more likely to perceive themselves as over-educated, it increases with 29.3 As a result hypothesis 4 cannot be rejected, although the evidence can only be

38

said with 90% security. When looking at model 4 is it clear that the increase of per-ceived over-education amongst respondent with a non-Dutch origin is attributed to the fact that non-Western immigrants are 51.3% more likely to perceive themselves as over-educated whereas the variable western-immigrants has no significant effect on perceived over-education. Therefore it can be found that non-western ethnicity seri-ously impacts perceived over-education amongst respondents. Hypothesis 5 is there-fore not entirely true because there is no significant effect when looking at western im-migrants. Yet it is the case that non-western immigrants are more likely to perceive themselves as over-educated, this can be said with 95% security.

Model 9 incorporates the variables gender, western immigrants, non-western immi-grants and the interaction effect between gender and the two immigrant categories. The control variables age and (semi-) public sector remain significant. Furthermore the only significant effect of the variable non-western immigrants on perceived over-educa-tion. According to the outcome, non-western immigrants are 1.790 times more likely to perceive themselves as over-educated. The remaining variables in the model show no significant effect. Consequently we have to reject hypothesis 6.

Model 7 includes the results of field of study and control variables. The results evi-dently show that field of study has no significant effect on perceived over-education. Again, the control variables age and (semi-) public sector demonstrate a significant ef-fect. Moreover, model 8 computes the gender and interaction variables of field of study with gender, these variables also don’t have a significant effect on perceived over-edu-cation. Hence, hypothesis 7 is also rejected. Model 10 includes the field of study varia-bles, the immigrant variable and the interaction variables immigrants and field of study. Interesting is that there is a significant effect within the interaction variable of immigrants and the study of humanities. The results exposes that immigrants with the field of study in humanities are 2.510 times more likely to perceive themselves as over-educated, although this can only be stated with a 10% significance. All the other variables in model 10 do not show a significant effect. Consequently hypothesis 8 is not entirely true because the study of humanities interacting with the variable immi-grants does show a significant positive effect on perceived over-education.

39