The relation between ethnic identity and field of study choice : a case study of the Netherlands

(1)

The Relation Between Ethnic Identity and Field

of Study Choice: A Case Study of the

Netherlands

Name: Hanna Mékdad Student Number: 10000529 Bachelor: Economics Supervisor: mw. S. He MSc

(2)

Abstract

The existence of wage differences across ethnic groups motivates researchers to study labour and education. This wage differential is partly driven by the differences between academic major choices of individuals (Porter and Umbach, 2006, 430). This encourages researchers to study the underlying factors that drive differences in academic major choices. There are several underlying factors such as: gender, personality and ethnicity. This study focuses on the relation between academic major choices and ethnicity. The study uses data from the LISS Panel, which is a survey conducted in the Netherlands. The hypotheses are tested using a multinomial logistic regression. The results show little significance of ethnic identity on academic major choice. This insignificance is driven by the lack of representativeness of the data. Another reason for the insignificant result is due to the exclusion of other factors that contribute to academic major choice. In summary, this study does not give a conclusive answer to whether ethnic identity influences academic major choices. This research therefore encourages future research on this subject.

(3)

I. Introduction

Labour is important for the production of goods and services. Education is further an essential component of labour, since it contributes to the labour force. There is a lot of literature on education, but there is little information on identity and education. The scarcity of literature provides a reason to further study the relation between identity and education. Identity plays an important role in decision-making according to the economists Akerlof and Kranton. Akerlof and Kranton infer that an individual’s sense of self, influences economic choices (Akerlof and Kranton, 2000). They extend their theory in an article about identity and schooling, where they find that individuals show specific behaviour that is compatible with a particular social

category (Akerlof and Kranton, 2002).

The current literature on the relation between ethnicity and academic major choice provides a foundation for further research. The focus of this thesis will be on how ethnicity affects academic major choice of an individual. Interest in this topic results from understanding how different wage differences across different ethnic groups occur. Research in this subject provides an explanation whether academic major choice differences between ethnic groups are the underlying source of differences in wages. This provides a further incentive for

policymakers to target differences in academic major choices that result in different wages across ethnic groups.

There are several studies about ethnic identity and academic major choice. The study of Porter and Umbach (2006) demonstrates that some ethnic groups are more likely to choose particular studies than other ethnic groups. The study of Maple and Stage (1991) focuses on ethnic identity and academic major choice. In this study different ethnic groups have different probabilities of choosing mathematics as their academic major. This analysis is supported by the empirical results of O’Brien, Martinez-Pons and Kopala (1991) who show that the

difference in self-confidence is the underlying factor that creates differences in academic major choices between different ethnic groups.

This study distinguishes itself from other studies because the empirical data is from Netherlands. The Netherlands differs from many aspects to the United States; there are different ethnicities and the schooling system is different. The employment of Dutch data shows whether the conclusion of American studies can be replicated to other countries such as the Netherlands.

(5)

from 2010. The use of a statistical model provides an answer to whether a relation is present between the dependent and independent variable. The thesis uses ethnicity as an independent variable and academic major choice as a dependent variable. The following section will focus on earlier empirical and theoretical studies. The third section includes the research method. The fourth chapters consist of results of the non-parametric and parametric tests. The discussion will be presented in the fifth chapter; this section provides an interpretation of the results with current literature and possible problems with the empirical findings. The final chapter provides a conclusion of this thesis.

The results of this thesis indicate that there is no significant relation between ethnic identity and academic major choice. There is little evidence according to the statistical tests to believe that ethnic identity influences differences in academic choices across ethnic groups. The insignificance of the results can be explained by the lack of the representativeness of the data. Further, the exclusion of other variables can attribute to the lack of significance of the results. The results of this study provide a reason to further research the relation between ethnic identity and academic major choice.

II. Background Information

A. Identity and Field of Study Choice

Identity is related to educational choices such as: effort and educational attainment (Akerlof and Kranton, 2002). According to Akerlof and Kranton individuals categorize themselves into social categories, which then in turn determine their behaviour. Individuals receive utility when they conform to the social group they belong to (Akerlof and Kranton, 2002, 1168). The

different ways in which people construct identity is by ethnicity, gender, socio-economic status, personality and political view (Porter and Umbach, 2006, 430).

Earlier studies show that there is a relation between gender and major choice in higher education (Maple and Stage 1991; Wilson and Boldizar 1990; Charles and Bradley 2002). The reasons for different academic majors due to gender are attributed by lack of confidence in mathematical studies of women or because of gender roles apparent in certain academic fields. According to several studies, women comprise a little part of science studies whereas men are still the majority that choose to do scientific or mathematical studies (Wilson and Boldizar, 1990, 63). The difference in gender distribution across fields of study is attributed by lack of confidence of women in their ability to do mathematics (Wilson and Boldizar, 1990, 63). Gender roles also cause women to choose different studies than men. An example of this

(6)

behaviour is the dominant presence of men in mathematical studies, since mathematics is perceived as a masculine field (Porter and Umbach, 2006, 431). Empirical results further show that in countries that have larger non-university sectors, women choose more female-dominant studies like social sciences and humanity studies (Charles and Bradley, 2002, 589).

In addition to gender, personality traits play a role in the choice an individual makes. According to empirical results, studies show that individuals with certain personality

characteristics are more likely to choose specific majors. Another empirical study also shows that career satisfaction and academic major choice are related. Feldman, Smart and Ethington (2004) show in an empirical article that personality and education choice are related. This is further supported by empirical evidence of Porter and Umbach (2006). People that are artistic are more likely to choose studies in the fields of art, music or languages (Porter and Umbach, 2006, 432). Holland’s theory is a model that is used to describe how personality and

environment influence behavioural choices. According to Holland’s theory individuals choose their environment based on their personality, thus individuals choose their environment based on compatibility with their personal character (Feldman, Smart and Ethington, 2004, 528). According to Feldman, Smart and Ethington (2004) students who choose a major that is not compatible with their personality does not necessarily mean they do not further develop their initial personality traits. Instead, students who choose a major that is not suitable with their personality develop other traits that are important for the chosen major (Feldman, Smart and Ethington, 2004, 542). However, studies do show that students who choose majors congruent with their personality are linked with higher education satisfaction and achievement (Porter and Umbach, 2006, 432). This indicates that individuals are better off when they choose an

academic major compatible with their personality.

B. Ethnicity and Field of Study Choice

There are not many theoretical explanations why different racial groups choose different majors. However, there are several studies that have empirical results showing that people with different racial backgrounds choose different academic majors.

The theoretical analysis of Akerlof and Kranton (2002) determines the relation between identity and choice that can be applied to ethnic identity and academic major choice. The use of game theory helps Akerlof and Kranton (2002) determine that students receive utility when they make behavioural choices that are conform to their self-image. The way in which

(7)

Kranton, 2002, 1168). This suggests that individuals from particular ethnic groups maximize their utility by choosing academic majors that are compatible with their ethnic identity. The model of Akerlof and Kranton implies that if a student deviates from his or her identity, a loss in self-image occurs. This happens because the model suggests that identity is a function of social category, ideal characteristics and behaviour of the specific social category. In the article they look at the effect of choosing specific social categories in high school and at the effort individuals exert. Whereas ei is one action of choice a student can choose from, one can replace

this ei with academic major choice. This results in a function where ethnic identity influences

the academic career choices of an individual. This academic major choice in turn leads to an individual being better off.

There are empirical studies that show the differences between ethnic groups and the academic career choice they make. Porter and Umbach (2006, 430) suggest that individuals with a specific ethnic background are less likely to choose a major where they are

underrepresented. According to the analysis black people are more likely to choose majors that are interdisciplinary rather than science majors compared to white people. Further they found that Hispanic people are more likely to choose humanity studies, art or social sciences rather than science majors compared to white people.

The article of Maple and Stage (1991) examines the factors that influence the relation between ethnicity and academic major choice. This study further examines the influence of parents on their child and the academic major choice of the child. Maple and Stage focus on how different ethnic groups have a higher probability to choose mathematics or science as their academic major. Maple and Stage are motivated by wage differences between minority groups and the underrepresentation of minorities in maths. The empirical results of the analysis show that the education of the mother and high school grades have positive effect on the probability an black individual chooses mathematics as a major (Maple and Stage, 1991, 51). The

prevalent factor that influenced the probability of a white individual choosing mathematics is high school grades (Maple and Stage, 1991, 51).

The research of O’Brien, Martinez-Pons and Kopala (1999) focuses on the underlying factors that influence different academic major choices across ethnic groups. The empirical results show that self-efficacy is an underlying factor that results in fewer minorities choosing scientific studies (O’Brien, Martinez-Pons and Kopala, 1999). According to other studies, self-efficacy is the self-confidence someone has to complete a certain task. The ethnic identity someone has, is associated with low or high self-efficacy. This can explain why there are low participation rates of minorities in mathematical studies (O’Brien, Martinez-Pons and Kopala,

(8)

1999, 231). The empirical results support the view that self-efficacy in mathematics/science determines the probability that an individual chooses a scientific or mathematical major. The study also shows that self-efficacy is a result of ethnic identity and high school academic performance, and therefore self-efficacy differs across different ethnic groups.

C. Hypotheses

The theoretical background provides a foundation for the hypotheses that are tested in this research. The first hypothesis is that: There are differences in the academic major choices between different ethnicity groups. This is based on the theory explained by Akerlof and Kranton (2002), Maple and Stage (1991) and O’Brien, Martinez-Pons and Kopala (1999). This study focuses on differences between ethnicities and not on the differences between Dutch people and non-Dutch people. The analysis provides an answer to whether an intrinsic difference is present between the academic choices of ethnic groups. This indicates that the study does not focus on the difference in choice between Dutch people and minorities.

The second hypothesis states that the proportion of other ethnicities is more likely to choose social or behavioural sciences compared to the proportion of Dutch people. The hypothesis test is applied to Turkish, Moroccan, Antillean, Surinamese, Indonesian, Non-Western people and Non-Western people. The foundation of the hypothesis results from the

empirical study of Porter and Umbach (2006). Porter and Umbach (2006) found that minorities, for example, Hispanics are more likely to choose social studies compared to Whites.

The third hypothesis is that Dutch people are more likely to choose science or mathematics as an academic major compared to other ethnic groups in the sample. This hypothesis is based on current literature (Maple and Stage, 1991). Science and social sciences are the two major fields in education, which indicates that all other fields are an extension of the major fields. This forms a possibility for the presence of overlap in the hypotheses. However, the three hypotheses do not overlap each other in this study. There are seven

different fields an individual can choose from in the survey. These seven different fields are not an extension of either science or social sciences. The seven different fields are mentioned in the next chapter of this thesis.

(9)

Hypothesis 1: H0: t1=m1=a1=s1=i1=nw1=w1=d1 H1: t1≠m1≠a1≠s1≠i1≠nw1≠w1≠d11 Hypothesis 2: H0: t1=d1 H1: t1>d1 H0: m1=d1 H1: m1>d1 H0: a1=d1 H1: a1>d1 H0: s1=d1 H1: s1>d1 H0: i1=d1 H1: i1>d1 H0: nw1=d1 H1: nw1>d1 H0: w1=d1 H1: w1>d1 Hypothesis 3: H0: t2=d2 H1: t2<d22 H0: m2=d2 H1: m2<d2 H0: a2=d2 H1: a2<d2 H0: s2=d2 H1: s2<d2 H0: i2=d2 H1: i2<d2 H0: nw2=d2 H1: nw2<d2 H0: w2=d2 H1: w2<d2

III. Conceptual Framework and Method

A. Sample and Data

The study uses data of the LISS (Longitudinal Internet Studies of Social Sciences) Panel, which is a large panel that consists of approximately 5000 households and 8000 individuals. The CentERdata (Tilburg University, The Netherlands) collects the data of the LISS Panel through a project called the MESS project. The panel consist of people representative to the Dutch population, and contains information on subjects including work, education, nationality and personality (LISS, 2014). The LISS panel is a longitudinal internet study that consists of six different points in time where information is collected from individuals. However, in this study only one wave is used for the statistical analysis. The use of one wave is done deliberately to make analysis easier and because this study does not focus on changes over time. The sample originates from the third wave. The data collection of this period is between 2010 and 2009.

B. Independent Variable

The independent variables that are included in this thesis are chosen on the basis of the

theoretical analysis in chapter two. Several factors influence academic major choice according to these articles. Some of these factors include personality, gender, self-efficacy and ethnicity. This thesis includes the variables that are mentioned by the empirical articles in chapter two.

1_t

1, m1, a1, s1, i1, nw1, w1 and d1 stand for the proportion of Turkish, Moroccan, Antillean, Surinamese, Indonesian, Non-Western or Western

people who choose behavioural or social sciences.

2 _t₂_{, m}₂_{, a}₂_{, s}₂_{, i}₂_{, nw}₂_{, w}₂_{and d}₂_{stand for the proportion of Turkish, Moroccan, Antillean, Surinamese, Indonesian, Non-Western or}

Western people who choose science/technology as a major.

(10)

In this study the variable ‘gender’ is a dummy variable, where the numerical value of being female is one and being male is zero. Several components of personality are given such as the variable creativity, which is an ordinal variable ranging from 1 for least creative and 7 for being highly creative. A creative person is more likely to choose an art major (Porter and Umbach, 2006, 433). The study further includes feelings of independence, logic, empathy, intellect, creativity and organization. The personality trait variables are all ordinal variables ranging from one to seven. Another factor that has been included is the feeling of confidence. According to the study of O’Brien, Martinez-Pons and Kopala (1999) self-confidence is positively correlated with choosing a study such as mathematics or science.

The main independent variable in this study is ethnicity. The idea behind this choice stems from the study of Porter and Umbach (2006). In the study of Porter and Umbach (2006) ethnicity is used as an independent variable. The study of Porter and Umbach includes three different ethnicity groups as dummy variables such as Whites, Blacks and Hispanics. In this study the variables of ethnicity are dummy variables that include ethnicities such as: Dutch, Moroccan, Turkish, Surinamese, Antillean, Indonesian, Western and Western. Non-Western and Non-Western are ethnicities of people who come from Non-Western or Non-Non-Western countries that are not specifically mentioned in the survey.

C. Dependent Variable

The dependent variable in this thesis is the field of study. In the questionnaire of the LISS panel, individuals can choose from 16 different academic majors. The academic major choices of an individual includes: general (no specific field), teaching, art, humanities,

social/behavioural studies, economics/business/administration/accountancy/management, law/public administration, mathematics/physics/IT, technology (which includes architecture, industry, crafts, construction etc.), medical studies (nursing, health services, etc.), personal care services (home economics, hair dressing school, etc.), catering/recreation, transport/logistics, telecommunication, public order/safety ( police, army, fire brigade etc.). The variables mentioned above are all categories each person can choose from in the survey. There are two different statistical tests employed in this research. The multinomial logistic regression includes only seven fields, since the inclusion of all 16 variables leads to problems in STATA (endless iterations). The solution for this is the combination of grouping several fields together in the regression model. The first field is art, the second field is social/humanities/behavioural

(11)

sciences, the third is economics/law, the fourth is science/math, the fifth is medical care, the sixth is practical studies and the seventh includes general studies and teaching.

D. Statistical Method

The methodology in this thesis consists of two parts; the first part is the non-parametric analysis. The second statistical method that is used in this study is a multinomial logistic regression model.

The non-parametric analysis consists of the calculation of the Cramer’s V correlation and the Chi-square test. These tests are applied to show whether variables are associated. The tests are applied for both ethnicity and gender. The non-parametric test includes the following fields: art, humanities, behavioural/social sciences, economics, math and technology.

The reason for the employment of the multinomial logistic regression model stems from the fact that the dependent variable (academic major choice) is a categorical variable that consists of more than four possibilities. The most appropriate model in the case of discrete nominal variables that are categories is the multinomial logistic regression model (Porter and Umbach, 2006, 440). In order to test whether some majors are more likely to be chosen by different racial groups, a base category is chosen. This is the same as what Porter and Umbach (2006) use in their study where they choose science as a base category. In this thesis ‘general field’ will be used as a base category since this category includes the majority of people of the sample.

IV. Results

A. Descriptive Statistics

This section provides a descriptive analysis that shows if there is an association between the dependent and independent variables. The association is tested with the use of the Cramer’s V correlation test and a Chi-square test. This section further presents the distribution of ethnic groups, gender and academic majors choice. The sample comprises of a total of 5,736 people, where 5532 are Dutch. The rest of the sample consists of Non-Dutch ethnic groups such as: Turkish, Moroccan, Antillean, Surinamese and Indonesian people. People of other Western and Non-Western countries are also included in the analysis. The exact numerical composition of ethnic groups is presented in table three in the appendix.

(12)

The first descriptive analysis of this study is the distribution of academic major choices given ethnicity and gender. The first statistical analysis provided in this section is the

percentage of Dutch and non-Dutch people choosing specific academic majors. The academic majors included in this analysis are art, humanities, social sciences, economics, maths and science or technology majors. The first graph shows the percentage of Dutch and non-Dutch people choosing different academic majors. The second graph shows the percentage of males and females choosing the specific academic majors mentioned before.

To provide a foundation for the statistical multinomial regression, a Cramer’s V test is conducted. The Cramer’s V is a statistical method that calculates the correlation between two categorical variables. The results of the Cramer’s V correlation can be found in table one for Dutch people, Non-Dutch people and gender. The result of the analysis shows a negative correlation between Dutch people and the choice of math or technology. Current literature shows that white people have a higher probability choosing mathematics or science as an academic major choice (Maple and Stage, 1991). The correlation non-Dutch people choosing maths or technology is however positive. The correlations provide an indication for a possible relation between ethnic groups and academic major choice. However, the correlation test show weak results between ethnicity and academic major choice. There is a comprehensive non-parametric analysis that is included in table five in the appendix. This is a table that shows the Cramer’s V and Chi-square results of all the different ethnicities in the sample.

Furthermore, the non-parametric analysis shows strong correlations between gender and academic major choice. The results show that there is a strong negative correlation between being female and choosing technology or mathematics. This result is consistent with the current literature. The current literature indicates that women are less likely to choose mathematics or science majors (Wilson and Boldizar, 1990, 63).

The second non-parametric method that is applied is the Chi-square test. The Chi-square test provides an additional measure to test whether a relation is present between academic major choice and gender and ethnicity (IDRE, 2014). The complete Chi-square test results are shown in table one. A significant Chi-square test shows that there is an association between ethnic identity and a certain academic major choice. Each ethnicity in the sample is measured by a dummy variable. The results show a negative association between being Dutch and choosing Humanities (p-value < 0.01). There is a positive association between Non-Dutch individuals and choosing Humanities as a major (p-value < 0.01). There is a significant negative association between Non-Dutch individuals and choosing social sciences (p-value <

(13)

A significant Chi-square test shows that there is a relation between being female and academic major choice. The variable gender shows strong associations between academic major choices compared to the variable ethnicity. The dummy variable gender equals one if an individual is a female. The results show a strong negative association between being female and choosing economics, mathematics or technology as an academic major (p-value < 0.01). There is a positive association between females and choosing art (p-value < 0.01).

TABLE 1. Correlation and association for ethnicity and academic major choice

Field of Study Cramer’s V Chi-square Test P-value

Dutch

Art 0.0010 0.0059 0.9390

Humanities -0.0367 7.6212*** 0.0060***

Social Sciences/Behavioural Studies 0.0200 2.2629 0.1330

Economics -0.0163 1.5084 0.2190 Math Technology -0.0146 -0.0200 1.2033 2.2698 0.2730 0.1320 Non-Dutch Art 0.0186 1.9469 0.1630 Humanities 0.0341 6.5732*** 0.0100*** Social Sciences -0.0244 2.8387* 0.0920* Economics 0.0067 0.2554 0.6130 Math 0.0141 1.1281 0.2880 Technology 0.0088 0.4385 0.5080 Gender Art 0.0353 7.0407*** 0.0080*** Humanities 0.0138 1.0732 0.3000

Social Sciences/Behavioural Studies 0.0590 19.6644*** 0.0000***

Economics -0.0421 9.9988*** 0.0020***

Math -0.1320 98.3681*** 0.0000***

Technology -0.3731 786.0979 0.0000***

Notes: Significance level measured by p-value * Significant at 10 percent level (p<0.1) ** Significant at 5 percent level (p<0.05) *** Significant at 1 percent level (p<0.001)

(14)

GRAPH 1. The percentage of individuals choosing specific academic majors for Dutch and Non-Dutch people

GRAPH 2. The percentage of individuals choosing specific academic majors for males and females 0 5 10 15 20 Art Humanities Social/Behavioural Studies Economics Math Technology Non-Dutch Dutch 0 5 10 15 20 25 30 35 Art Humanities Social/Behavioural Studies Economics Math Technology Male Female

(15)

B. Multinomial Logistic Regression

The study uses multinomial logistic regression to see if ethnicity and academic major choice are related. There are several academic major choices in the sample, but to make statistical regression easier several academic majors are grouped in categories. This limits the amount of iterations in STATA. The academic major category now consists of: art, social sciences (social sciences, behavioural sciences and humanities), economics/law, science (mathematics and technology), medical care, practical (transport, personal care, catering, telecommunication, agriculture and public order) and general (teaching, no specific field). The use of three different models of logistic regression shows that the addition of control variables has a remarkable effect on the coefficients of ethnicity.

The first multinomial logistic regression model only includes the dependent and independent variable. The results show that the model is insignificant (p-value=0.3053 > 0.100). The significance of a model shows whether the model is useful in predicting the probability of choosing a specific field given the independent variable (IDRE, 2014). The pseudo-R2 of the model is 0.0030. The pseudo-R2 in the multinomial logistic regression is calculated with the use of the formula of McFadden. A higher pseudo-R2 indicates that the model is a better fit, but it is not the same as the normal R2 of OLS (Laitila, 1993). The model shows that Dutch and individuals from other Western countries are less likely than other ethnic groups to choose social sciences. There are other ethnic groups that have a lower probability than Dutch or Western people to choose social or humanities studies. However, these results are not significant.

The second model shows that the addition of gender results in a significant model (p-value=0.000 < 0.001). This indicates that the model is able to significantly predict the

probability of choosing a certain field given ethnicity. The explanatory power of the model is given by pseudo-R2 that now is higher than the previous model (pseudo-R2 =0733). This result shows that adding gender as a control variable improves the predictive power of the model. Most of the coefficients are insignificant in this model, but the models shows that being Western has a significant positive effect on the probability of choosing art. According to this model Dutch people have a significant negative effect on the probability of choosing social sciences. Individuals from other Western countries have a negative probability choosing social sciences. Furthermore, people from other Western countries are less likely to choose social sciences as their academic major compared to Dutch people. This is because the coefficient of the Western variable is smaller than the coefficient of the Dutch ethnicity.

(16)

The third model adds personality as a control variable; this choice is based on the theoretical analysis given in the second chapter. The model is significant and has a p-value of 0.000. The model further has a pseudo-R2 of 0.0898. This indicates that adding personality improves the fit of the model. The model does not show many coefficients that are significant. The model does show that individuals from other Western countries besides the Netherlands have a positive probability of choosing mathematics or science. The results do not show that other ethnic groups have a significant effect on the probabilities of choosing specific academic majors. There are control variables that do have a significant effect on the probability of choosing certain majors. People who are more creative have a significant positive effect on the probability of choosing social sciences. Creativity does not significantly affect the probability of an individual choosing art in the model. However, Porter and Umbach (2006, 432) do show that creative people are more likely to choose artistic majors. Further, creative people or females are less likely to choose economics or law. Being a female also reduces significantly the possibility of choosing mathematics or science. Females are also less likely to choose practical studies compared to men. The non-parametric comparison of this result can be found in table five in the appendix. People who score themselves high on empathy and intellect show a significant negative effect on the probability of choosing a practical major. The coefficients of the multinomial logistic regression can be found in table two.

TABLE 2. The results of the Multinomial Logistic Regression coefficients.

Field of Study Model 1 (N=5206) Model 2 (N=5206) Model 3 (N=975)

Art Dutch 0.6974 0.7038 0.1756 Turkish -13.6645 -13.9948 -16.0861 Moroccan -14.4676 -14.7821 -16.6874 Antillean -17.7316 -20.2343 -1.34841 Surinamese -15.4308 -15.7074 -17.0168 Indonesian -15.2514 -15.6831 0.0000 Western 1.4688 1.4454* -15.5968 Non-Western 1.6592 1.6615 -15.1182 Gender 0.0926 0.9965

(17)

Empathy -1.1853 Organization 15.4181 Self-confidence -0.3325 Independence -1.3891 Intellect 14.5208 Logic 14.0857 Creativity 14.9591 Social Sciences Dutch -2.1211* -2.0992* -27.9228 Turkish -1.7363 -1.7440 -27.3170 Moroccan -1.1792 -1.1205 -44.0644 Antillean 1.5102 1.4781 8. 3435 Surinamese -18.2933 -18.5538 -45.84199 Indonesian -17.1977 -17.5794 0.0000 Western -2.8404** -2.8198** -43.3637 Non-Western -0.9721 -1.0054 -25.7748 Gender 0.1137 0.3810 Empathy 0.0842 Organization 0.1327 Self-confidence 0.1538 Independence 0.4965 Intellect 0.0914 Logic -1.3098** Creativity 1.6048** Economics/Law Dutch -0.1258 -0.1533 -0.1743 Turkish -0.1098 -0.2158 -1.0007 Moroccan -0.5326 -0.4963 0.1729

(18)

Antillean -17.9315 -20.0948 0.3756 Surinamese -16.2872 -16.5662 -18.4746 Indonesian -15.9427 -16.0687 0.0000 Western 0.2147 0.2739 -0.2151 Non-Western 0.2135 0.2030 1.1188 Gender -0.6402*** -0.6749*** Empathy -0.0640 Organization 0.0188 Self-confidence -0.1416 Independence 0.1774 Intellect -0.1488 Logic -0.0192 Creativity -0.6749** Science Dutch -0.2652 -0.3676 0.5167 Turkish 0.2003 -0.1293 -0.9264 Moroccan -0.6062 -0.4901 -16.1825 Antillean -17.9391 -18.4042 1.6995 Surinamese -16.4221 -16.2429 -18.3283 Indonesian -16.0051 -14.4922 0.0000 Western 0.1625 0.4006 1.5504* Non-Western 0.2932 0.3067 1.2156 Gender -2.8970*** -2.6025*** Empathy -0.0265 Organization 0.0681 Self-confidence -0.5458 Independence 0.0069 Intellect -0.5531

(19)

Logic 0.2065 Creativity -0.0368 Medical Care Dutch 0.6275 0.6319 0.3886 Turkish -0.2101 -0.0127 0.2683 Moroccan -14.7372 -14.9511 -16.4156 Antillean 0.4080 0.1156 21.7339 Surinamese -15.5447 -15.7358 -16.3822 Indonesian -15.5977 -16.2856 0.0000 Western 0.5396 0.4353 0.3302 Non-Western -0.7642 -0.7187 0.4585 Gender 1.2629*** 1.5042*** Empathy -0.1435 Organization 0.0156 Self-confidence 0.9709 Independence -0.4325 Intellect 0.2071 Logic -0.4504 Creativity -0.4105 Practical Dutch -0.1311 -0.1627 0.6150 Turkish -0.6424 -0.7726 0.1902 Moroccan -14.9626 -15.2467 -16.5943 Antillean 0.9115 1.2716 -0.3875 Surinamese -16.3074 -16.5661 -17.6507 Indonesian 0.8543 1.2011 0.0000 Western -0.2690 -0.1905 -0.3255 Non-Western -1.2172 -1.2251 -15.3310

(20)

Gender -0.8041*** -0.3688* Empathy -0.4630* Organization 0.2215 Self-confidence 0.4405 Independence 0.5087 Intellect -0.6225* Logic -0.2368 Creativity 0.0365

Notes: * Significant at 10 percent level (p<0.1). ** Significant at 5 percent level (p<0.05) *** Significant at 1 percent level (p<0.001)

The variable Indonesian is omitted in model 3 due to collinearity.

C. Conclusion

In summary, the results show that ethnicity has no significant effect on the academic major choice of an individual. However, there are some significant results, but these do not remain significant after personality and gender are added as control variables. The model does show that gender has a significant effect on several academic major choices. This indicates that being a female or male does have an effect on academic major choice. The same applies to

personality, having certain personality traits contribute to a higher or lower probability of choosing an academic major choice. Further, adding control variables results in a model that has more explanatory power. Finally, the results indicate that the null hypotheses stated in chapter two cannot be rejected, since they are not significantly different from zero.

V. Discussion

A. Limitations

There are possible limitations in this study that provide an explanation for the insignificant results. First, the number of other ethnicities besides Dutch is limited in the sample. In the sample 5532 people have a Dutch origin and 204 people have another ethnic origin. This

(21)

Bureau of Statistics, the population consists of approximately 3.5 million non-Dutch and 13.2 million Dutch people (CBS, 2014) 3. These numbers indicate that almost 25 percent of the Dutch population has another ethnic origin besides Dutch. In the sample approximately 3.5 percent has another ethnicity than Dutch.

Second, the sample does not include all variables that influence academic major choice, such as parental background. According to an article of Dustman (2004) secondary track choice is related to parental background. The LISS panel does not include parental background as a variable. The omission of parental background can lead to insignificant coefficients. It is further advisable for future research to control for local effect and non-local effect. The aim of this study is namely to analyse the relation between different ethnicities and academic major choice. However, being an immigrant results in different choices compared to non-immigrants. It is therefore advised for future research to control for this local and non-local effect. This study used the LISS panel that comprises of people who have a Dutch passport. There is no evidence that the people in the sample are first or second generation immigrants. There is a separate immigrant panel available at the CentERdata.

A possible solution for future research on ethnic identity and education choice is the use of data that is representative for the Dutch population. Furthermore, adding control variables, such as parental background, lead to a better statistical model. However, there are still many other variables that affect academic major choice. It is advised for future research to include more variables and to choose a dataset that is representative of the population. The

methodology of this study shows that with the representative data, the results are more significant. The addition of gender in the non-parametric analysis shows more significant results compared to ethnicity. This result is also present in the multinomial logistic regression, where gender has a significant effect on the probability of choosing certain academic majors. The distribution of gender is more representable for the Dutch population compared to ethnicity variable. The percentage of women in the sample is approximately 60 percent and for the males it is 40 percent. This indicates that with the use of representative data, the methodology that is used in this study is capable of analysing the relation between ethnicity and academic major choice.

3_{The exact number of the population in the Netherlands is: 13,234,545 Dutch people and 3,594,744 non-Dutch}

people. The numbers are based on an analysis from 2014 and the percentages are calculated by the author of this thesis

(22)

B. Interpretation of Results

The results in this study show no significant effect of being from a specific ethnic group and academic major choice. This is tested by the Cramer’s V correlation, the Chi-square test and the multinomial logistic regression.

The non-parametric results show weak correlations between ethnic identity and

academic major choice, the results all range around 10 percent. This shows that ethnic identity does not significantly influence academic major choice. The results of the correlation are further supported by the results of the Chi-square tests. There are three different combination of ethnicity and academic major choice that are significant. The first one is being Dutch and choosing humanities. There is a negative Cramer’s V that indicates that Dutch people are less likely to choose humanities as a major. The Chi-square test complements this, because it shows a negative association between the two variables. The second significant result is that Non-Dutch people are more likely to choose humanities. Results of the Cramer’s V are positive and the chi-square test is significantly positive. However, Non-Dutch people are less likely to choose social/behavioural sciences, this is shown by a negative Cramer’s V of approximately two percent. This is further supported by the Chi-square test that is significant. There is a stronger relation between gender and academic major choice compared to ethnicity. According to the results female individuals are more likely to choose art and social/behavioural sciences. The positive Cramer’s V and Chi-square test show that there is a positive association. The non-parametric analysis further shows that there is a strong negative association between female individuals choosing economics, mathematics and technology. This shows that females are less likely than men to choose scientific or economic majors. This result is consistent with the background literature in chapter two.

The multinomial logistic model does not show a significant link between ethnic identity and the probability of choosing a certain academic major. The results are not consistent across the different models. This can be seen when all control variables are added. The coefficient of people from Western countries and choosing social sciences is not significant in the third model. The model contains little evidence that indicates that ethnic identity is related to different academic major choice. According to the first multinomial logistic model, Dutch individuals are less likely to choose social/behavioural studies compared to a general study. General studies are the base category in the multinomial logistic regression model. The same answer applies to individuals from other Western countries besides the Netherlands. In the

(23)

The same applies again to people from other Western countries. In the third model people from other Western countries besides the Netherlands are more likely to choose science. The third model does not further show a significant link between ethnic identity and academic major choice when all the control variables are added.

In the multinomial logistic regression model the addition of control variables leads to a less significant link between ethnic identity and academic major choice. However, several control variables show that they have a significant effect of an individual choosing a particular major. Females are less likely to choose science or mathematics compared to general studies. This result is congruent to previous literature, indicating that females are less likely to choose mathematics (Wilson and Boldizar, 1990, 63). The same goes for women and practical academic majors. Women are less likely to choose practical studies such as public order or agriculture as an academic major. According to the third model, personality traits of an individual do have a significant effect on the probability of choosing certain majors. Creative people are more likely to choose social sciences compared to general sciences. According to empirical research of Porter and Umbach (2006, 432) creative people are also more likely to choose art, but in this study the result is not significantly different from zero. Creative people are also less likely to choose economics or law as an academic major. Logical individuals are less likely to choose social sciences as a major. Individuals who see themselves as intellectual are less likely to choose practical majors such as public order, personal care or agriculture. These results are consistent with Holland’s theory of personality and academic major choice. People that are intellectual are less likely to choose practical studies, whereas creative people are less likely to choose studies such as economics (Porter and Umbach, 2006, 433).

The hypotheses that are presented in the second chapter cannot be rejected according to the statistical analyses. According to the results of the Chi-square test there is no significant relation between ethnicity and the choice of an academic major. The results of the multinomial logistic regression indicate that the first null hypothesis cannot be rejected. The second

hypothesis states that the proportion of ethnic groups is more likely to choose art or

social/behavioural sciences compared to Dutch people. This null hypothesis cannot be rejected, because the results of the multinomial logistic regression do not show many significant

coefficients. This indicates that the coefficients of the model are not significantly different from zero. The same applies to the third hypothesis that states that Dutch individuals are more likely to choose scientific majors compared to other ethnic groups. The third null hypothesis cannot be rejected, because the multinomial logistic regression coefficients are not significantly different from zero. Possible reasons for the insignificant results are given in the previous

(24)

section. This study cannot give a conclusive answer on the research question. This means that according to this study there is no reason to believe that ethnic identity results in different academic major choices across ethnic groups in the Netherlands.

VI. Conclusion

The focus of this study is the relation between ethnic identity and academic major choice. Research on this subject is of importance since it explains wage differences across ethnic groups. There are several factors that influence individuals to choose specific academic majors. The first influential factor is gender; according to several studies females choose differently compared to men. Females tend to shy away from mathematical or scientific fields. Moreover, personality is also a factor that contributes to different choices of individuals. According to Holland’s personality theory, individuals choose an environment compatible with their personality. This theory can be further applied to academic major choice. Ethnicity and academic major choice are according to current literature related. However there is not a conclusive answer to how ethnic identity influences individuals to choose different academic majors.

This study focuses on the question how ethnic identity and academic major choice are related. This study focuses on information of Dutch data. However after several statistical analyses there is no conclusive answer to the research question. The results that are presented in the previous chapters are insignificant. According to this study ethnic identity does not influence academic major choice. However, the results do show that gender and personality characteristics influence academic major choice.

According to this research ethnic identity and academic major choices of an individual are not related. The reason for the insignificant results can be attributed by the lack of

representativeness of the data. The data that is used in this study does not include a high number of other ethnic people besides Dutch. Furthermore, there are several factors that

influence academic major choice but are not included in this study. The limitations of this study stimulate the need for further researching the relation of ethnic identity and academic major choice.

(25)

VII. Reference List

Akerlof, George A., Rachel E. Kranton (2000). Economics and Identity.The Quarterly Journal of

Economics 115 (3): 715–53.

Akerlof, George A., Rachel E. Kranton (2002). Identity and Schooling: Some Lessons for the Economics of Education. Journal of Economic Literature 40 (4): 1167–1201.

CBS (2014). Bevolking; Generatie, Geslacht, Leeftijd and Herkomstgroepering. Accessed on July 5 2014.

http://statline.cbs.nl/StatWeb/publication/?VW=T&DM=SLNL&PA=37325&D1=0&D2

=a&D3=0&D4=0&D5=2-4,11,38,46,95- 96,137,152,178,182,199,220,237&D6=0,4,8,12,16,l&HD=140707-1607&HDR=T,G2,G3,G5&STB=G1,G4

Charles, Maria, Karen Bradley (2002). Equal But Seperate ? A Crossnational Study of Sex Segregation in Higher Education.”American Sociological Review Vol. 67 (May 2014): 573–99. http://www.jstor.org/stable/3088946.

Dustmann, Christian. (2004). Parental Background, Secondary School Track Choice, and Wages.

Oxford Economic Papers 56 (2): 209–30. doi:10.1093/oep/gpf048.

Feldman, Kenneth A., John C. Smart and Corinna A. Ethington (2004). What Do College Students Have To Lose? Exploring the Outcomes of Differences in Person-Environment Fits. The Journal of Higher Education 75 (5): 528–55.

http://www.jstor.org/stable/3838762.

IDRE (2014). What Statistical Analysis Should I Use? Accessed on July 5 2014. http://www.ats.ucla.edu/STAT/stata/whatstat/default.htm

Laitila, Thomas. (1993). A pseudo-R2 measure for limited and qualitative dependent variable models. Journal of Econometrics. 56(3): 341-355.

http://www.sciencedirect.com/science/article/pii/030440769390125O

LISS (2014). The Representativeness of LISS, an Online Probability Panel. Accessed on May 21. https://www.lisspanel.nl/assets/uploaded/representativeness_LISS_panel.pdf Maple, Sue A., Frances K. Stage (1991). Influences on the Choice of Math/Science Major by

Gender and Ethnicity.” American Educational Research Journal 28 (May 2014): 37–60. http://www.jstor.org/stable/1162878 .

O’Brien, Virginia, Manuel Martinez-pons and Mary Kopala (1999). Mathematics Self-Efficacy, Ethnic Identity, Gender, and Career Interests Related to Mathematics and Science. The

Journal of Educational Research 92 (4): 231–35. doi: 10.1080/00220679909597600.

Porter, Stephen R., Paul D. Umbach (2006). College Major Choice: An Analysis of Person-Environment Fit. Research in Higher Education 47 (4): 429–49. doi:10.1007/s11162-005-9002-3.

Wilson, Kenneth L., Janet P. Boldizar (1990). Gender Segregation in Higher Education: Effects of Aspirations, Mathematics Achievement and Income. Sociology of Education 63 (1): 62–74. http://www.jstor.org/stable/2112897.

(26)

VIII. Appendix.

TABLE 3. Number of ethnic people and gender

Group Number of People

Dutch Non Dutch 5532 204 Turkish 32 Moroccan 13 Antillean 8 Surinamese Indonesian

Other Non-Western Country Other Western Country Female Male 3 2 28 118 3032 20154

(27)

GRAPH 3. The percentage of individuals choosing specific academic majors for different ethnic groups.

TABLE 4. The correlation and association for ethnicity and academic major choice.

Dutch

Art 0.0010 0.0059 0.9390

Humanities -0.0367 7.6212*** 0.0060***

Social Sciences/Behavioural Studies 0.0200 2.2629 0.1330

Economics -0.0163 1.5084 0.2190 Math Technology -0.0146 -0.0200 1.2033 2.2698 0.2730 0.1320 Turkish Art -0.0751 1.5010 0.2840 Humanities -0.0270 0.1490 0.6990

Social Sciences/Behavioural Studies Economics Math Technology -0.0610 -0.0844 -0.0270 -0.0065 0.7591 1.4529 0.1490 0.0085 0.3840 0.2280 0.6990 0.9260 Moroccan Art -0.0454 0.4208 0.5170 Humanities 0.1394 3.9643** 0.0460**

Social Sciences/Behavioural Studies -0.0369 0.2777 0.5980

Economics -0.0090 0.0164 0.8980 Math Technology -0.0560 -0.0056 0.6408 0.0064 0.4250 0.9360 0 5 10 15 20 25 30 Art Humanities Social/Behavioural Studies Economics Math Technology Western Non-Western Antillean Moroccan Turkish Dutch

(28)

Antillean

Art -0.0352 0.2523 0.6150

Humanities -0.0434 0.3843 0.5350

Social Sciences/Behavioural Studies 0.3357 22.9916*** 0.0000***

Economics -0.0904 1.6653 0.1970 Math Technology -0.0434 -0.0880 0.3843 1.6069 0.5350 0.2050 Surinamese Art -0.0213 0.0923 0.7610 Humanities -0.0262 0.1405 0.7080

Economics -0.0546 0.6090 0.4350 Math Technology -0.0262 -0.0537 0.1405 0.5876 0.7080 0.4430 Indonesian Art -0.0173 0.0612 0.8050 Humanities -0.0214 0.0932 0.7600

Social Sciences/Behavioural Studies -0.0141 0.0404 0.8410 Economics Math -0.0445 -0.0214 0.4040 0.0932 0.5250 0.7600 Technology -0.0437 0.3898 0.5320 Non-Western Art 0.0992 2.0071 0.1570 Humanities 0.0530 0.5740 0.4490

Social Sciences/Behavioural Studies -0.0564 0.6491 0.4200 Economics Math Technology 0.0127 0.1224 -0.0592 0.0331 3.0569* 0.7141 0.8560 0.0800* 0.3980 Western Art 0.0294 0.1766 0.6740 Humanities -0.0605 0.7473 0.3870

Economics 0.1112 2.5210 0.1120

Math 0.0363 0.2690 0.6040

Technology 0.1013 2.0916 0.1480

Notes: Significance level measured p-value * Significant at 10 percent level (p<0.1) ** Significant at 5 percent level (p<0.05) *** Significant at 1 percent level (p<0.001)

(29)

TABLE 5. The correlation and association for gender and practical academic major choices.

Gender

Catering 0.0356 7.1640*** 0.0070***

Transport -0.1282 92.8233*** 0.0000***

Public Order -0.1106 69.0710*** 0.0000***

Personal Care -0.1356 103.8421*** 0.0000***

Notes: Significance level measured by p-value * Significant at 10 percent level (p<0.1) ** Significant at 5 percent level (p<0.05) *** Significant at 1 percent level (p<0.001)

The relation between ethnic identity and field of study choice : a case study of the Netherlands