Ethnic and Gender Penalties
at the Dutch Labor Market:
the role of different forms of capital
for first generation immigrants from
Morocco and Turkey
Abstract
The aim of this research is to answer the following research questions: (1) can differences in human capital, social capital and cultural capital explain ethnic penalties in employment and occupational status of first generation immigrants from Turkey and Morocco as compared to Dutch natives in the Netherlands? And (2) can differences in human capital, social capital and
cultural capital explain both ethnic and gender penalties in employment and occupational status of first generation Turkish and Moroccan women? In order to answer these questions
data of the Netherlands Longitudinal Lifecourse Study (NELLS) (de Graaf, Kalmijn, Kraaykamp & Monden, 2010) has been analyzed by running multivariate regressions. The
results show that human capital is the most important predictor of employment and occupational status. More specifically, educational level is most important for employment and language proficiency for occupational status. The results also show that first generation
immigrants from Morocco and Turkey experience ethnic penalties on employment and occupational status. For employment, these are partly explained by the different forms of
capital together. For occupational status, the forms of capital together explain the ethnic penalty in total. Again, human capital is the most important for explaining the ethnic penalty.
Lastly, the results show that Moroccan and Turkish women do not experience ethnic and gender penalties on employment and only experience a gender penalty on occupational status.
This gender penalty is explained by all forms of capital together and human capital is again the most important for explaining this penalty.
Ilse Zoon (10780904) 26-06-2018 Bachelor thesis:
Economic Integration and Disadvantage of Immigrants First reader: Agnieszka Kanas
Second reader: Matthijs Kalmijn University of Amsterdam
Content
Introduction ... 1 Theoretical framework ... 3 Human capital ... 3 Social capital ... 4 Cultural capital ... 4 Ethnic penalties ... 5 Double disadvantage ... 6Data and methods ... 8
Data ... 8 Measurement ... 10 Dependent variables ... 10 Independent variables ... 11 Method ... 13 Results ... 14 Descriptive analysis ... 14 Human capital ... 14 Social capital ... 15 Cultural capital ... 16
Human-, social-, and cultural capital compared ... 17
Ethnic penalties ... 18
Double disadvantage ... 21
Conclusions and discussion ... 24
References ... 28
Table 1: Descriptive statistics ... 31
Table 2: Logistic regression on employment ... 32
Table 3: Logistic regression on employment with gender ... 33
Table 4: OLS regression on occupational status ... 34
1
Introduction
The integration of immigrants at the Dutch labor market has been subject of much research. Compared to natives they experience more difficulties in finding a job (Niesing, van Praag & Veenman, 1994; Crul & Doomernik, 2003), are more often underemployed (Zorlu & Hartog, 2008; Falcke, 2017), and have jobs with a lower socioeconomic status (Gracia, Vazquez & van de Werfhorst, 2014; Kanas & van Tubergen, 2009). The disadvantage of immigrants at the labor market can be referred to as an ethnic penalty (Falcke, 2017; Kalter & Kogan, 2006). Additionally, immigrant women are mostly more disadvantaged than men, in other words, they also experience a gender penalty. Therefore, previous research speaks of a double disadvantage for immigrant women (Boyd, 1984). For instance, employment rates in the Netherlands are lower for Moroccan and Turkish immigrants than for natives and the employment rates of women are lower than that of men (Centraal Bureau voor de Statistiek StatLine, 2018).
In general it is argued that immigrants have less resources than natives in order to succeed at the labor market and experience ethnic penalties because of this. Different forms of capital are seen as important resources in order to succeed at the labor market. Firstly, human capital is believed to be of positive influence on labor market outcomes because this raises one’s productivity and efficiency (Becker, 1975). For immigrants in specific is believed that host country specific human capital is more valuable than country of origin human capital because of problems with transferability and differences in quality (van Tubergen & van de Werfhorst, 2007). Secondly, social capital is believed to be of positive influence on labor market outcomes because it provides for information and influence on the job-matching process (Kanas & van Tubergen, 2009). For immigrants in specific, the contacts with natives are the most valuable because they can provide for new and more valuable information and influence on the Dutch labor market (Lancee, 2012; Kanas & van Tubergen, 2009). Thirdly, cultural capital is believed to be of positive influence on labor market outcomes because being common with the cultural signals of the high status culture reduces the chance of social
exclusion (de Graaf, de Graaf & Kraaykamp, 2000). First generation Moroccan and Turkish immigrants, who generally come from low social backgrounds (Crul & Doomernik, 2003), are expected to have low levels of cultural capital which in turn lowers their chance to succeed at the Dutch labor market. Previous studies examined whether these forms of capital separately could explain ethnic penalties and concluded that immigrants experience ethnic penalties, even after taking these forms of capital into account (Zorlu & Hartog, 2008; Gracia, Vazquez
2
& van de Werfhorst, 2014; Crul & Doomernik, 2003; Falcke, 2017, Niesing, van Praag & Veenman, 1994). However, these studies have not examined if the three forms of capital together can explain ethnic penalties in labor market outcomes. Also, it remains unclear whether the forms of capital can explain the double disadvantage of immigrant women.
Two important indicators of labor market outcomes are one’s employment status and one’s occupational status. First of all, employment status is a basic indicator of labor market outcomes since it indicates whether one has a job to provide for himself/herself (Lancee, 2012). When looking at employment, there can be differentiated between the active and inactive labor force participation. The active labor force participation consists of those who are employed or actively searching for a job whereas the inactive labor force participation consists of those who are unemployed and not searching for a job (Lancee, 2012). Previous studies have mostly examined employment status without differentiating between an active and inactive labor force participation (Gracia, Vazquez & van de Werfhorst, 2014; Kanas & van Tubergen, 2009; Lancee, 2012). It is argued that it is not necessary to do this because this measures the likelihood of finding employment regardless of one’s situation (Lancee, 2012). However, this makes it unclear whether the predictors influence one’s decision to work or one’s likelihood to get hired after this decision is made. Therefore, this research will only measure employment status for the active labor force participation: those who are not actively looking for a job are not subject of this research. Secondly, differences exist within
employment because occupations differ in the income and prestige or power they provide. For this reason it is useful to measure occupational status as well, since this indicates the income as well as the prestige or power that is associated with an occupation (Lancee, 2012).
The aim is to answer the following research questions: (1) can differences in human capital, social capital and cultural capital explain ethnic penalties in employment and
occupational status of first generation immigrants from Turkey and Morocco as compared to Dutch natives in the Netherlands? And (2) can differences in human capital, social capital and cultural capital explain both ethnic and gender penalties in employment and occupational status of first generation Turkish and Moroccan women? There is focused on first generation immigrants from Turkey and Morocco because the integration of Turkish and Moroccan immigrants is considered the most problematic among the largest non-Western immigrant groups in the Netherlands: Turks, Moroccans, Surinamese, Dutch Antilleans and Dutch Eurasians (Crul & Doomernik, 2003).
3
Theoretical framework
Human capital
Human capital can be defined as the capacity of an individual to be productive and efficient. It consists of knowledge, skills, and health. A combination of this influences someone’s
productivity (Becker, 1975). From this theoretical perspective, employers would rationally hire employees based on their human capital (Gracia, Vazquez & van de Werfhorst, 2014). Thus, according to human capital theory, investments in human capital are of positive influence on labor market outcomes (Becker, 1975). Previous research on human capital of immigrants has measured human capital mainly by looking at educational attainment and language proficiency (Gracia, Vazquez & van de Werfhorst, 2014). In other words, a higher level of education and better language skills would be more valuable on the labor market than a low level of education and few language skills.
For human capital of first generation immigrants it is necessary to differentiate
between host country specific human capital and country of origin human capital (Bevelander & Groeneveld, 2012). Knowledge and skills are not perfectly mobile across countries
(Chiswick, 1978) so human capital investments outside the host country may not be transferable to the host country (van Tubergen & van de Werfhorst, 2007). Also, previous research found that there are large differences in the quality of human capital across countries (Kaarsen, 2011). Human capital obtained in the Netherlands is presumed to be more valuable than human capital obtained in Turkey and Morocco (van Tubergen & van de Werfhorst, 2007; Kanas & van Tubergen, 2009). According to this, education obtained in the Netherlands and a better understanding of the Dutch language in specific would be more valuable on the Dutch labor market than education abroad and a good understanding of another language.
Hypothesis 1: Immigrants who have a higher level of education, obtained education in the Netherlands and speak the Dutch language well are more likely to be employed and have a higher occupational status as compared to those with lower levels of education, did not obtain education in the Netherlands, and speak the Dutch language less well.
4
Social capital
Coleman has defined social capital as a variety of entities that (1) consist of social structures, and (2) facilitate actions of actors within that structure (Coleman, 1988). Thus, social capital theory is concerned with the connectedness or engagement of an actor with other individuals, organizations, or communities (Brucker, 2015) and the actions that are influenced by this. According to social capital theory, social capital affects labor market outcomes because social relations provide valuable resources (Portes & Sensenbrenner, 1993; Lancee, 2012). These resources are mostly defined as information and influence. Firstly, information on which jobs are available, how to look for jobs in general, how to present oneself in an interview and, eventually, how to behave on the job. Secondly, influence on the job-matching process by providing an entry into occupations (Kanas & van Tubergen, 2009). Based on the access to host country specific resources for the labor market, immigrants can be categorized as a resource-poor group and natives as a resource-rich group (Lancee, 2012). This implies that contacts with natives are very valuable for immigrants because they firstly provide new forms of information and influence, and secondly, because these forms of information and influence are more valuable on the Dutch labor market than the information and influence out of the contacts with other immigrants (Lancee, 2012; Kanas & van Tubergen, 2009). There is assumed that immigrants have less contact with natives than natives do which lowers their chances to succeed at the labor market.
Hypothesis 2: Immigrants who have more contact with natives are more likely to be employed and have a higher occupational status as compared to those who have less contact with natives.
Cultural capital
The basis of cultural capital theory is that cultural capital is a tool for dominant status groups and social classes to maintain their power and it is a way for parents of the elite group to make sure that their children are advantaged as compared to the children of the non-elite. Cultural capital consists of widely shared, high-status cultural signals that are used for social and cultural exclusion. Cultural signals are, for example, behaviors, tastes, and attitudes. Cultural capital theory mostly refers to the importance of socialization into certain high-status
activities. Such as an interest in art, listening to classical music, attending the theater and museums, and reading literature. One’s cultural capital is thus of a certain level because of the
5
cultural capital of one’s parents. According to cultural capital theory, an individual who is not common with the cultural signals of the high-status culture will have a higher chance of social and cultural exclusion (de Graaf, de Graaf & Kraaykamp, 2000). This makes it harder for people who come from lower social backgrounds to succeed at the labor market than for people who come from higher social backgrounds: they lack the skills, habits, and styles that are rewarded at the labor market (Gracia, Vazquez & van de Werfhorst, 2014). Moroccan and Turkish immigrants generally come from lower social backgrounds than the average Dutch native (Crul & Doomernik, 2003) so it is expected that they less frequently participate in high-status activities. Which in turn lowers their chance to succeed at the Dutch labor market.
Hypothesis 3: Immigrants who more frequently attend classical concerts, opera or ballets, historic museums, art museums, and the theater are more likely to be employed and have a higher occupational status as compared to those who less frequently attend classical concerts, opera or ballets, historic museums, art museums, and the theater.
Ethnic penalties
As stated, a disadvantage on labor market outcomes for ethnic minorities as compared to natives is commonly called an ethnic penalty: a penalty because of their ethnic attributes (Kalter & Kogan, 2006; Gracia, Vazquez & van de Werfhorst, 2014). This ethnic penalty could be the source of a lack of resources that are necessary to succeed (Kalter & Kogan, 2006). As already stated, first generation immigrants of Morocco and Turkey are believed to have less valuable human capital, less contact with natives, and do less frequently participate in high-status activities than natives. This lack of resources could explain the ethnic penalty on labor market outcomes for Moroccans and Turks. However, it is impossible to control for everything that is of influence on labor market outcomes so there is believed that a lack of resources can only partly explain the ethnic penalty (Kalter & Kogan, 2006). Previous research on ethnic penalties has also shown that controlling for the different forms of capital separately does partly explain the ethnic penalty but part remains unexplained (Zorlu & Hartog, 2008; Gracia, Vazquez & van de Werfhorst, 2014; Crul & Doomernik, 2003; Falcke, 2017, Niesing, van Praag & Veenman, 1994).
Hypothesis 4: Turks and Moroccans experience ethnic penalties at employment and occupational status as compared to Dutch natives.
6
Hypothesis 5: The ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their human capital as measured by the level of highest educational attainment, Dutch educational credentials and language proficiency.
Hypothesis 6: The ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their social capital as measured by their social ties with natives.
Hypothesis 7: The ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their cultural capital as measured by the frequency of attending classical concerts, opera or ballets, historic museums, art museums, and the theater.
Hypothesis 8: The ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their human-, social-, and cultural capital together.
Double disadvantage
For immigrant women there can be spoken of a double disadvantage. Firstly, because of their ethnicity and secondly, because of their gender (Boyd, 1984). Most research has examined this by looking at the choice to work: Turkish and Moroccan women are believed to exclude themselves from the labor market due to traditional gender role attitudes (Khoudja &
Fleischmann, 2014; Bevelander & Groeneveld, 2012). Traditional gender role attitudes are expected to negatively influence labor market outcomes of women because they pressure women to prioritize domestic work above paid work (Khoudja & Fleischmann, 2014; Corrigal & Conrad, 2007). First generation immigrant women from Morocco and Turkey are expected to come from an environment with more traditional gender role attitudes (Khoudja &
Fleischmann, 2014) which could explain their double disadvantage at the labor market. However, when looking at women who chose to be active on the labor market, previous research has shown that women are still disadvantaged as compared to men (Macedo & Santos, 2013). Traditional gender role attitudes cannot explain this disadvantage. According
7
to the theories on human-, social-, and cultural capital, labor market outcomes are determined by these forms of capital. Based on these theories it is expected that differences in these forms of capital can account for the gender penalty of Moroccan and Turkish women as compared to Moroccan and Turkish men and the ethnic penalty as compared to Dutch women and thus explain these penalties. However, previous research has also shown that equal opportunities do not guarantee equality of outcome for men and women (Macedo & Santos, 2013). It is thus expected that human-, social-, and cultural capital can only partly explain the penalties.
Hypothesis 9: Turkish and Moroccan women experience an ethnic penalty as compared to Dutch women and a gender penalty as compared to their male counterparts at employment and occupational status.
Hypothesis 10: The penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their human capital as measured by the level of highest educational attainment, Dutch educational credentials and language proficiency.
Hypothesis 11: The penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their social capital as
measured by their social ties with natives.
Hypothesis 12: The penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their cultural capital as
measured by the frequency of attending classical concerts, opera or ballets, historic museums, art museums, and the theater.
Hypothesis 13: The penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their human-, social-, and cultural capital together.
8
Data and methods
Data
This research uses data from the first wave of the Netherlands Longitudinal Lifecourse Study (NELLS) (de Graaf, Kalmijn, Kraaykamp & Monden, 2010). This dataset contains cross-sectional data which have been conducted in two phases: (1) from December 2008 until July 2009 and (2) from October 2009 through May 2010. The research population consists of Turkish and Moroccan immigrants in the Netherlands and a reference group that consists of inhabitants that are not Turkish or Moroccan with an age of 15 to 45. One’s ethnicity is determined by the country of birth of one’s mother or, when the mother is born in the Netherlands, by the country of birth of one’s father. A respondent is defined as Dutch when both of the parents are born in the Netherlands (de Graaf, Kalmijn, Kraaykamp & Monden, 2010).
Sampling has been done by two-stage stratified sampling: (1) a quasi-random selection of 35 municipalities by region and urbanization, and (2) a random selection from the
population registry by age and ethnicity. The municipalities were sampled stratified by three regions (West, North/East and South) and four degrees of urbanization based on the amount of addresses per square kilometer: (1) very strong (2500 or more addresses), (2) strong (1500 to 2500 addresses), (3) moderate (1000 to 1500 addresses), and (4/5) Marginal/not urbanized (less than 1000 addresses) (Centraal Bureau voor de Statistiek, 2018). The selecting of municipalities was not completely random because the four big cities in the West (Amsterdam, Rotterdam, Den Haag and Utrecht) had to be included in order to have a representative sample of Moroccans and Turks. After selecting the municipalities, local authorities were asked to draw three random samples from the population registry: inhabitants with an age of 15 to 45 who were (1) born in Morocco or whose father or mother was born in Morocco, (2) born in Turkey or whose father or mother was born in Turkey, and (3)
inhabitants with an age of 15 to 45 excluding those belonging to group (1) and (2). This led to a total sample of 12310 addresses. In the end, 5312 respondents participated of whom 1143 Turks with a response rate of 50%; 1192 Moroccans with a response rate of 46%; and 2977 others with a response rate of 56% (de Graaf et al., 2010). The group of others consists of 227 non-Western immigrants, 229 Western immigrants, and 2556 Dutch natives (de Graaf et al., 2010).
9
The data have been conducted by using a questionnaire that consists of two parts: a face-to-face interview and a self-completion questionnaire. Before the fieldwork, the interviews were tested among 100 Turks, 100 Moroccans, and 100 other inhabitants of the Netherlands. During the first phase of the fieldwork the interviewer started with the face-to-face interview and left the self-completion questionnaire with the respondent which could also be completed on the Internet. Initially, because of this approach, many respondents did not fill in the self-completion questionnaire. Therefore, during the last months of the first phase and the second stage another approach was used. The respondents were asked to fill in the self-completion questionnaire on the Internet before the interview and when they did not do so the interviewer waited until the respondent finished the self-completion questionnaire after the face-to-face interview. Still, 7.7% of the sample has not completed the self-completion questionnaire (De Graaf et al., 2010).
This dataset is useful for this research because it contains data on the two biggest non-Western immigrant groups in the Netherlands, as well as on Dutch natives. The survey also contains relevant questions regarding human capital, social capital, cultural capital, and labor market outcomes. One disadvantage of using this dataset is that the data was conducted from 2008 up until 2010, so it can be questioned if this is still relevant. However, statistics of
Centraal Bureau voor de Statistiek show that the penalties remain an issue, for example,
employment rates of women are 64% for natives, 48% for Moroccans and 48.3% for Turks whereas employment rates of men are 72.6% for natives, 65% for Moroccans and 73.2% for Turks (Centraal Bureau voor de Statistiek StatLine, 2018).Another disadvantage is that not all municipalities are randomly selected which lowers the representativeness of the sample and in turn the external validity of this research. Also, the response rate is relatively low for Moroccans and Turks as compared to the reference group which can cause a bias which will not be accounted for by weighted variables. A disadvantage of cross-sectional data in general is that it is impossible to examine causality. There is examined whether there is a relation and based on theory is assumed that this relation is not spurious but there cannot be examined that this is no reverse causality (Lancee, 2012). For example, there cannot be concluded that a significant effect for social capital means that social capital influences labor market outcomes since it can also be the other way around: that one’s position on the labor market influences one’s social capital (Brucker, 2015).
For the purpose of this research the total sample of the NELLS dataset is reduced to a research population that consists of first generation immigrants from Turkey and Morocco, and Dutch natives. First generation immigrants are those who are born in Turkey or Morocco
10
and whose parents are not both Dutch. Dutch natives are those who are born in the
Netherlands and whose parents are also born in the Netherlands (Centraal Bureau voor de Statistiek, 2012; de Graaf et al., 2010). There is focused on respondents who are economically active, that is, those who work for at least twelve hours a week, or are unemployed but search for a job (Centraal Bureau voor de Statistiek, 2018). The amount of hours worked per week is determined by the actual amount of hours worked, not by the amount of hours that is in the contract. Whether a respondent is searching for a job is measured by two items: (1) whether the respondent is looking for a job at the moment, and (2) how many job interviews the respondent has had in the last three months. Whenever a respondent scores yes on the first item or more than 0 on the second item this is coded as searching for a job. Also, people who are still in education are excluded. This leaves a sample of 2678 respondents of whom 480 are Turkish (17.9%), 435 are Moroccan (16.2%), and 1763 are Dutch (65.8%). After excluding the respondents with missing values on at least one of the independent variables the sample is reduced by 7.8% to an amount of 2470 respondents of whom 402 are Turkish (16.3%), 361 are Moroccan (14.6%), and 1707 are Dutch (69.1%).
Measurement
Dependent variables
The first dependent variable is employment. Respondents were asked whether they were active in paid labor or not at the time of the interview. Self-employment is included in employment. This has been recoded so the variable indicates whether one is (1) employed or (0) unemployed but looking for employment.
The second dependent variable is occupational status. The dataset includes a variable that indicates one’s current occupation in terms of the ISCO-88 scale. This has first been transformed to the ISCO-08 scale and in turn to the ISEI-08 scale. The International Socio-Economic Index (ISEI) measures the hierarchical position of an occupation that is linked to education and income. A higher score indicates a higher occupational status (Ganzeboom, De Graaf & Treiman 1992). This variable has a range from 11.56 to 88.7.
11
Independent variables
Ethnicity
The first three independent variables are dummy variables based on ethnicity: Turkish, Moroccan and Dutch. They are based on one item that indicates one’s ethnic origin based on the self-reported country of birth of the respondent and both parents of the respondent. This has been recoded so the variable Turkish indicates whether one is a first generation immigrant from Turkey (1) or not (0). The variable Moroccan indicates whether one is a first generation immigrant from Morocco (1) or not (0). And the variable Dutch indicates whether one is a Dutch native (1) or not (0).
Gender
One’s gender is indicated by the dummy variable female. This variable is based on the sex that is registered in the municipal register. This has been recoded so that it indicates whether one is female (1) or male (0).
Human capital
Human capital has been measured by education and language proficiency. Education is measured by the independent variables education and Dutch education. These variables are measured by three items. First, the respondents were asked which levels of education they had followed. The answer categories are based on the Dutch schooling system and contain the following categories from low to high: no education, primary school, lower general secondary education (VMBO BB or KB), preparatory secondary vocational education (VMBO GL or TL), higher general secondary education (HAVO), pre-university education (VWO), short primary lower vocational education (MBO level 1/2), medium/long tertiary/secondary lower vocational education (MBO level 3/4), higher vocational education (HBO), university bachelor, university master/doctoral, and PhD. Additionally, there were three categories for forms of education that are not comparable to the levels of Dutch education: primary-, secondary-, and tertiary education. The dataset includes dummy variables on the different categories that indicate whether the respondent followed the level of education or not. Secondly, the respondents were asked whether they obtained a diploma for each level of education they followed. The dataset includes dummy variables on every level of education that indicate whether one obtained a diploma of the given level. Thirdly, the respondents were asked where they followed each level of education. The dataset includes variables on every level of education that indicate where one followed the given level with the following
12
categories: in the Netherlands, not in the Netherlands, and partly in the Netherlands and partly not. For the purpose of this research, not comparable primary education is coded as primary education, not comparable secondary education as HAVO, and not comparable tertiary education as university bachelor. The categories university master/doctoral, and PhD are merged into one category because the two categories are too small when taken separately. The answer category that indicates that one followed a level of education partly in the Netherlands and partly not is interpreted as Dutch education because these respondents followed at least one part within the Netherlands, which means they obtained host-country specific human capital. Firstly, the variables on the first two items are recoded into the variable education so that they indicate the highest diploma of education, regardless of the country where the diploma was obtained. This variable has a range from (0) no education to (10) university master/doctoral/PhD. Secondly, a dummy variable Dutch education is created that indicates whether one has a diploma that is obtained within the Netherlands (1) or not (0). In order to create this variable, the variables on the three items have been recoded into a variable that indicates one’s highest diploma of education within the Netherlands. Whenever one scores higher than 0, so when one has a diploma obtained within the Netherlands, one scores 1 at Dutch education. When one has no diploma of education at all or no diploma obtained within the Netherlands Dutch education is coded as 0.
Language proficiency is measured by four items indicating how well one can (1) understand Dutch spoken language, (2) speak Dutch, (3) read in Dutch, and (4) write in Dutch. The answers range from (1) very well to (5) not. This has been recoded so that the variables have a range from (0) not to (4) very well. For Dutch natives is assumed that they score very well on all four items. These variables are recoded into one variable language proficiency (Cronbach’s Alpha=0.97). This variable has a range from 0 to 4.
Social capital
Social capital is measured by the frequency of contact with natives and whether one has Dutch friends. The frequency of contact with natives consists of native neighbors and natives in unions or clubs. Respondents were asked respectively how often they had contact with (1) native neighbors, and (2) with natives within unions or clubs. The answers range from (1) (almost) every day to (7) never. Additionally, there is an eighth category for respondent’s who are not a member of a union or club. This is coded as never. These variables have been
recoded so that a higher score indicates a higher frequency of contact. The variables have a range from 0 to 6 and could not be computed into one scale (Cronbach’s Alpha=0.29). Dutch
13
friends is measured by one item that indicates whether one has a good friend of every of the following descents: (1) Dutch, (2) Turkish, (3) Moroccan, (4) Surinamese or Antillean, and (5) other Non-western. The dataset includes dummy variables that indicate whether one has a friend of the given ethnic descent. This is recoded into a dummy variable that equals 1 for those who have Dutch friends and 0 for those who do not.
Cultural capital
Cultural capital consists of classical concert, opera or ballet, historic museum, art museum, and theater. This has been measured by one item that indicates how often the respondent attended to each of the following high-cultural activities in the twelve months before the interview: (1) classical concerts, operas or ballets, (2) historic museums, (3) art museums, and (4) theaters. The answer categories range from (1) never to (5) 12 or more times. This has been recoded so that the variables have a range from (0) never to (4) 12 or more times. These variables could not be computed into one scale (Cronbach’s Alpha=0.62).
Method
In order to analyze the data and test the hypotheses, four multivariate regressions are applied. The choice for the regressions is based on the measurement scale of the dependent variable. Two logistic regressions are applied on a dichotomous variable for employment and two Ordinary Least Squares (OLS) regressions are applied on occupational status. The models of the logistic regressions are similar to the models of the OLS regressions. Every regression consists of five models. For both employment and occupational status the first regression includes variables in the following order: model 1 includes the variables on ethnicity (Moroccan, Turkish and Dutch), model 2 includes variables on ethnicity and human capital (education, Dutch education and language proficiency), model 3 includes variables on
ethnicity and social capital (Dutch neighbors, Dutch natives within unions or clubs, and Dutch friends), model 4 includes variables on ethnicity and cultural capital (classical concert, ballet, or opera, historic museum, art museum, theater), and lastly, model 5 includes variables on ethnicity and all three forms of capital. The second regression for both employment and occupational status includes the variables in the same order but additionally includes gender (female) and interaction variables of female*Moroccan and female*Turkish in every model. Dutch is the reference category for every model of every regression and every model of every regression is controlled for age.
14
Before running the regressions was checked whether the data meets the assumptions that logistic and OLS regression are based on. Linearity, homoscedasticity, and mean
independence are checked for by running a linear regression model for both dependent variables with all independent variables and requesting a scatterplot for the standardized residuals with the standardized predicted value. For all tables the assumptions of linearity, homoscedasticity, and mean independence are met. Whether there is multicollinearity is checked for by running a linear regression model for both dependent variables with the Z-scores of all independent variables and requesting collinearity diagnostics in order to check if the errors are correlated in a problematic way. The condition index of all variables is below 15 so the assumption of uncorrelated errors is met: there is no problem with collinearity. Normal distribution of errors is checked for by requesting histograms on variables on the standardized residuals that were saved while requesting the scatterplot. For all regressions the assumption of a normal distribution of errors is met.
Results
Descriptive analysis
Table 1 shows the descriptive statistics of all dependent and independent variables. Interesting to note is that 14% of Moroccans and Turks is involuntarily unemployed as opposed to 3% of the Dutch natives. Also, the mean score on occupational status is 36.86 for Moroccans, 38.36 for Turks, and 49.16 for Dutch natives on a scale of 11.56 to 88.7. This already indicates that first generation immigrants from Morocco and Turkey are disadvantaged on employment and occupational status as compared to Dutch natives. Table 1 also shows that the mean score on all independent variables concerning human-, social-, and cultural capital is lower for
Moroccans and Turks than for Dutch natives. This indicates that first generation immigrants from Morocco and Turkey have less resources in order to succeed at the Dutch labor market than Dutch natives.
Human capital
Table 2 and 4 show the regressions independent of gender. M2 of table 2 and 4 includes ethnicity and human capital in the regression (education, Dutch education and language proficiency). The hypothesis concerning human capital is that immigrants who have a higher level of education, obtained education in the Netherlands and speak the Dutch language well are more likely to be employed and have a higher occupational status as compared to those
15
with lower levels of education, did not obtain education in the Netherlands, and speak the Dutch language less well (H1). In line with this hypothesis, M2 of table 2 shows positive effects for educational level and language proficiency on employment. Only the effect of educational level is significant. Not in line with the hypothesis is a negative but not significant effect for having a Dutch diploma. In line with the hypothesis, M2 of table 5 shows
significant positive effects for educational level and language proficiency on occupational status. Not in line with the hypothesis, M2 of table 5 shows a negative but not significant effect for having a Dutch diploma. The odds of being employed versus unemployed are 1.26 (p-value < 0.001) for educational level, 0.66 (p-value > 0.05) for having a Dutch diploma, and 1.19 (p-value > 0.05) for language proficiency. The B-coefficient on occupational status is 3.54 (p-value < 0.001) for educational level, -2.59 (p-value > 0.05) for having a Dutch diploma, and 5.64 (p-value < 0.001) for language proficiency. This also shows that educational level is the most valuable predictor of human capital for one’s chances to be employed. While this is still an important predictor for occupational status language
proficiency is a more valuable predictor than educational level for occupational status. For H1 can thus be concluded that having obtained one higher level of education raises the likelihood of being employed by (1.26-1)*100=26% as compared to one lower level. Having a Dutch diploma and scoring one level higher on language proficiency does not significantly raise the likelihood of being employed as compared to not having a Dutch diploma and scoring one level lower on language proficiency. The score on occupational status is raised by 3.54 points when one has one higher level of education as compared to one lower level and by 5.64 points when one scores one level higher on language proficiency as compared to one level lower. Having a Dutch diploma does not significantly raise one’s score on occupational status as compared to not having a Dutch diploma.
Social capital
M3 of table 2 and 4 shows the regressions independent of gender when including ethnicity and social capital. The hypothesis concerning social capital is that immigrants who have more contact with natives are more likely to be employed and have a higher occupational status as compared to those with less contact with natives (H2). In line with this hypothesis, M3 of table 2 shows positive effects on employment for all variables on social capital but only the effect for the frequency of contact with natives in unions or clubs is significant. Also in line with the hypothesis, M3 of table 5 shows significant positive effects on occupational status for
16
the frequency of contact with natives in unions or clubs and having Dutch friends. Not in line with the hypothesis is a significant negative effect on occupational status for the frequency of contact with native neighbors. For employment, the odds of being employed versus
unemployed are 1.08 (p-value > 0.05) for the frequency of contact with native neighbors, 1.08 value < 0.05) for the frequency of contact with natives in unions or clubs, and 1.35 value > 0.05) for having a Dutch friend. The B-coefficient on occupational status is -0.92 (p-value < 0.001) for the frequency of contact with native neighbors, 0.75 (p-(p-value < 0.001) for the frequency of contact with natives in unions or clubs, and 9.06 (p-value < 0.001) for having a Dutch friend. This also shows that having more contact with natives in unions or clubs is the most valuable predictor for the likelihood to be hired in paid employment, while having a Dutch friend is the most valuable predictor for occupational status. Having more contact with Dutch neighbors is the least valuable for occupational status. It even is a negative predictor. For H2 can thus be concluded that one higher frequency of contact with natives in unions or clubs raises the likelihood of being employed by (1.08-1)*100=8% as compared to one lower frequency of contact with natives in unions or clubs. One higher frequency of contact with native neighbors and having a Dutch friend does not significantly raise the likelihood of being employed as compared to one lower frequency of contact with native neighbors and not having a Dutch friend. The score on occupational status is 0.92 points lower when having one higher frequency of contact with native neighbors as compared to one lower frequency, 0.75 points higher when having one higher frequency of contact with natives in unions or clubs as compared to one lower frequency, and 9.06 points higher when having a Dutch friend as compared to not having a Dutch friend.
Cultural capital
M4 of table 2 and 4 shows the regressions independent of gender when including ethnicity and cultural capital. The hypothesis with regard to cultural capital is that immigrants who more frequently attend classical concerts, operas or ballets, historic museums, art museums, and the theater are more likely to be employed and have a higher occupational status as compared to those who less frequently attend classical concerts, operas or ballets, historic museums, art museums, and the theater (H3). In line with this hypothesis, M4 of table 2 shows positive effects on employment for the frequency of attending a classical concert, opera or ballet, a historic museum, and the theater but only the effect for attending the theater is significant. Not in line with the hypothesis is a negative but not significant effect on
17
employment for attending an art museum. In line with the hypothesis, M4 of table 4 shows significant positive effects for all forms of cultural capital. The odds of being employed versus unemployed are 1.48 (p-value > 0.05) for the frequency of attending a classical concert, ballet, or opera, 1.35 (p-value > 0.05) for the frequency of attending a historic museum, 0.71 value > 0.05) for the frequency of attending an art museum, and 1.53 (p-value < 0.05) for the frequency of attending the theater. The B-coefficient on occupational status is respectively 4.35 (p-value < 0.001), 2.56 (p-value < 0.001), 4.01 (p-value < 0.001), and 3.86 (p-value < 0.001). This also shows that attending the theater is the most valuable predictor of cultural capital for employment. For occupational status the order from most valuable predictor to least valuable predictor is: attending a classical concert, opera, or ballet, attending an art museum, attending the theater, and last, attending a historic museum. For H3 can thus be concluded that one higher frequency of attending the theater raises the likelihood of being employed by (1.53-1)*100=53% as compared to one lower frequency. One higher frequency of attending a classical concert, ballet, or opera, of attending a historic museum, and of attending an art museum does not significantly raise the likelihood to of being
employed as compared to one lower frequency. One higher frequency of attending a classical concert, ballet or opera, of attending a historic museum, of attending an art museum, and of attending the theater as compared to one lower frequency raises one’s score on occupational status by respectively 4.35, 2.56, 4.01, and 3.86 points.
Human-, social-, and cultural capital compared
When comparing M2 to M4, table 2 shows that for every form of capital there is only one aspect that significantly raises the likelihood of being employed. For human capital this is the level of educational attainment, for social capital this is the frequency of contact with natives in unions or clubs, and for cultural capital this is the frequency of attending the theater. Of these three forms, attending the theater raises the likelihood of being employed the most, this is followed by level of education, and the frequency of contact with neighbors raises the likelihood of being employed the least. Table 4 shows that the only aspects that do not significantly raise one’s score on occupational status are having a Dutch diploma for human capital and the frequency of contact with native neighbors for social capital. It also shows that having a Dutch friend raises one’s score on occupational status the most, which is followed by language proficiency, and third comes the frequency of a attending a classical concert, ballet, or opera. M5 includes ethnicity together with every form of capital. M5 of table 2 shows that
18
the significant effect on employment of educational level is reduced by 0.01 odds to 1.25 (p-value < 0.001). The significant effect of the frequency of contact with native neighbors and attending the theater is not significant anymore and the frequency of attending an art museum shows a significant negative effect on employment (odds: 0.63; p-value < 0.05). In other words, when controlling for the other forms of capital only human capital remains
significantly positive. More specifically, one higher level of education does significantly raise the likelihood of being employed by (1.25-1)*100=25% as compared to one lower level. Interestingly, one higher frequency of attending an art museum lowers the likelihood of being employed by (0.63-1)*100=37% as compared to one lower frequency. The other aspects are no significant predictors for the likelihood of being employed. M5 of table 4 shows that the only significant effects that become not significant when controlling for the other forms of capital are the positive effects on the frequency of contact with natives in unions or clubs and the frequency of attending a historic museum. All other significant effects are reduced but remain significant. In M5 language proficiency raises one’s score on occupational status the most instead of having a Dutch friend, this is now followed by educational level, and having a Dutch friend comes only third.
Ethnic penalties
M1 of table 2 and 4 includes ethnicity in the regression independent of gender. The
hypothesis regarding M1 of these tables is: Turks and Moroccans experience ethnic penalties at employment and occupational status as compared to Dutch natives (H4). In line with this hypothesis, M1 of table 2 and 4 show significant negative results for Turks and Moroccans at employment and occupational status. The odds to be employed versus unemployed as
compared to natives are 0.21 for Moroccans (p-value < 0.001) and 0.2 for Turks (p-value < 0.001). The B-coefficient on occupational status is -12.32 for Moroccans (p-value < 0.001) and -10.92 for Turks (p-value < 0.001). H4 can thus be confirmed: Turks and Moroccans experience ethnic penalties at employment and occupational status. The likelihood of being employed for Moroccans and Turks is respectively (0.21-1)*100=79% and (0.2-1)*100=80% lower than that of natives. On occupational status Moroccans and Turks score respectively 12.33 and 10.92 points less than natives. The ethnic penalty on employment is slightly bigger for Turks than for Moroccans, while the ethnic penalty on occupational status is slightly bigger for Moroccans than for Turks.
19
M2 to M4 all include ethnicity together with one form of capital in the regression. M2 includes ethnicity and human capital, M3 includes ethnicity and social capital, and M4
includes ethnicity and cultural capital. Firstly, the hypothesis regarding M2 of table 2 and 4 is that the ethnic penalties experienced by Turks and Moroccans at employment and
occupational status as compared to Dutch natives are partly explained by differences in their human capital as measured by the level of highest educational attainment, Dutch educational credentials and language proficiency (H5). Secondly, the hypothesis regarding M3 of the same tables is that the ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their social capital as measured by their social ties with natives (H6). Thirdly, the hypothesis regarding M4 is that the ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their cultural capital as measured by the frequency of attending classical concerts, operas or ballets, historic museums, art museums, and the theater (H7). In line with these hypotheses, M2 to M4 of table 2 and 4 show that the ethnic penalties of Turks and Moroccans on
employment and occupational status as compared to Dutch natives are reduced as compared to M1. Most effects are still significant which indicates that the forms of capital separately do not explain the total of ethnic penalties. Only the effect of Turks on occupational status becomes not significant when controlling for human capital. M2 to M4 also show that human capital reduces the ethnic penalties the most, this is followed by cultural capital, and lastly, social capital reduces the ethnic penalties the least. The odds of being employed versus unemployed of Moroccans as compared to natives were 0.21 in M1 (p-value < 0.001). Controlling for human capital raises these odds by 57.1% to 0.33 (p-value < 0.001), controlling for cultural capital raises the odds by 28.6% to 0.27 (p-value < 0.001), and controlling for social capital raises the odds by 23.8% to 0.26 (p-value < 0.001). The odds of being employed versus unemployed of Turks as compared to natives were 0.2 in M1 (p-value < 0.001). Controlling for human capital raises these odds by 50% to 0.3 (p-value < 0.001), and both controlling for cultural capital and social capital raises the odds by 20% to 0.24 (p-value < 0.001). In all models the ethnic penalty on employment of Turks is still bigger than that of Moroccans. The negative B-coefficient on occupational status of Moroccans as
compared to natives was -12.32 in M1 (p-value < 0.001). When controlling for human capital the negative B-coefficient is reduced by 80.4% to -2.42 (p-value < 0.05), when controlling for cultural capital this is reduced by 31.3% to -8.47 (p-value < 0.001), and when controlling for social capital this is reduced by 18.7% to -10.02 (p-value < 0.001). The negative B-coefficient
20
on occupational status of Turks as compared to natives was -10.92 in M1 (p-value < 0.001). When controlling for human capital this B-coefficient becomes positive (0.23; p-value > 0.05). This effect is not significant. Cultural capital reduces the negative B-coefficient by 27.7% to -7.89 (p-value < 0.001), and social capital by 23% to -8.41 (p-value < 0.001). In all models the ethnic penalty on occupational status of Moroccans is still bigger than that of Turks. In other words, H5 to H7 can be confirmed. All forms of capital partly explain the ethnic penalty of Turks and Moroccans as compared to Dutch natives on employment and occupational status. Human capital even explains the total ethnic penalty of Turks on occupational status. The likelihood to be hired was respectively 79% and 80% lower for Moroccans and Turks than for natives in M1. This is only (0.33-1)*100=67% and (0.3-1)*100=70% when controlling for human capital, (0.27-1)*100=73% and (0.24-1)*100=76% when controlling for cultural capital, and (0.26-1)*100=74% and (0.24-1)*100=76% when controlling for social capital. The score on occupational status was respectively 12.32 and 10.92 points lower for Moroccans and Turks than for natives in M1. For Moroccans this is only 2.42 points when controlling for human capital and Turks do not have a significant lower score than natives when controlling for human capital. When controlling for cultural capital Moroccans and Turks still respectively score 8.47 and 7.89 points lower than natives, and when controlling for social capital this is 10.02 and 8.41 points. Table 2 and 4 also show that the M2 has the largest explained variance. The explained variance in the logistic regression on employment is measured in terms of the Nagelkerke R-square. This is 0.16 when including human capital and 0.11 for both social capital and cultural capital. The explained variance in the OLS regression on occupational status is 32% when including human capital, 9% when including social capital, and 15% when including cultural capital.
M5 includes ethnicity and all forms of capital in the regression. The hypothesis concerning M5 of table 2 and 4 is that the ethnic penalties experienced by Turks and Moroccans at employment and occupational status as compared to Dutch natives are partly explained by differences in their human-, social-, and cultural capital together (H8). In line with this hypothesis, M5 of table 2 shows that the ethnic penalties of Moroccans and Turks on employment are reduced and that these results are still significant. For occupational status the results are not significant anymore, the three forms of capital together can explain the total ethnic penalty on occupational status of Moroccans and Turks as compared to natives. There is also shown that when controlling for all forms of capital together the ethnic penalties are reduced more than when only controlling for one of the three forms of capital. The odds of being employed versus unemployed as compared to Dutch natives are respectively raised by
21
71.4% and 60% for Moroccans and Turks as compared to M1. The new odds are 0.36 (p-value < 0.001) for Moroccans (M1: 0.21; M2: 0.33; M3: 0.26; M4: 0.27) and 0.32 (p-(p-value < 0.001) for Turks (M1: 0.2; M2: 0.3; M3: 0.24; M4: 0.24). The negative B-coefficient on occupational status of Moroccans and Turks as compared to natives is reduced by 95.5% for Moroccans and becomes positive for Turks. The new B-coefficients are -0.56 (p-value > 0.05) for Moroccans and 1.66 (p-value > 0.05) for Turks. Both effects are not significant. In other words, H8 can be confirmed for employment: the three forms of capital together do partly explain the ethnic penalties on employment for Moroccans and Turks as compared to natives. The likelihood to be hired for paid employment for Moroccans is still (0.36-1)*100=64% lower than that of natives and for Turks this is even (0.32-1)*100=68% lower than that of natives. Thus, a large part of the ethnic penalties of Moroccans and Turks on employment remains unexplained. The three forms of capital together do explain the total of ethnic
penalties for Moroccans and Turks on occupational status. The effects of being Moroccan and Turkish are not significant anymore so both Turks and Moroccans do not significantly have a lower score on occupational status than natives when taking human-, social-, and cultural capital into account. Table 2 and 4 also show that human capital is probably the most important form of capital in explaining ethnic penalties. The Nagelkerke R-square of M5 of table 2 is only 0.01 point higher than when only including human capital in M2: 0.17 as compared to 0.16. For M5 of table 4 the adjusted R-square is 2% higher than M2: 34% as compared to 32%. Thus, adding social- and cultural capital does not influence the explained variance much.
Double disadvantage
Table 3 and 5 shows the regressions on employment and occupational status while including gender. M1 only includes ethnicity and gender. The hypothesis regarding M1 of these tables is: Turkish and Moroccan women experience an ethnic penalty as compared to Dutch women and a gender penalty as compared to their male counterparts at employment and occupational status (H9). Not in line with this hypothesis, M1 of table 3 shows no significant effects for the interaction variables female*Moroccan and female*Turkish. For H9 can thus be concluded that the likelihood of being employed is not significantly lower for Turkish and Moroccan women as compared to native women as well as to their male counterparts. In line with the hypothesis is that M1 of table 5 shows significant effects for the interaction variables. The B-coefficient on occupational status is 7.4 value < 0.01) for female*Moroccan and 7.42
(p-22
value < 0.01) for female*Turkish. In other words, there is a significant difference in
occupational status for Moroccan and Turkish women as compared to native women and their male counterparts. The size of this difference is -15.12+7.4= -7.72 for Moroccan women as compared to Moroccan men, 13.85+7.42= -6.4 for Turkish women as compared to Turkish men, -1.07+7.4= 6.33 for Moroccan women as compared to native women, and -1.07+7.42= 6.35 for Turkish women as compared to native women. In line with H9 can thus be concluded that the score on occupational status is 7.72 points lower for Moroccan women then for Moroccan men, and 6.4 points lower for Turkish women as compared to Turkish men. Not in line with H9 can be concluded that the score on occupational status is 6.33 points higher for Moroccan women as compared to native women, and 6.35 points higher for Turkish women as compared to native women. Overall, Moroccan and Turkish women do not experience a double disadvantage on employment and occupational status. On employment they experience no penalties and on occupational status they only experience a gender penalty.
M2 to 4 all include ethnicity and gender together with one form of capital in the regression. M2 includes ethnicity, gender and human capital, M3 includes ethnicity, gender and social capital, and M4 includes ethnicity, gender and cultural capital. The hypothesis regarding M2 is that the penalties experienced by Turkish and Moroccan women at
employment and occupational status are partly explained by differences in their human capital as measured by the level of highest educational attainment, Dutch educational credentials and language proficiency (H10). For M3 the hypothesis is that the penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their social capital as measured by their social ties with natives (H11). Lastly, the hypothesis for M4 is that the penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their cultural capital as measured by the frequency of attending classical concerts, operas or ballets, historic museums, art museums, and the theater (H12). Since M1 of table 3 has already shown that Moroccan and Turkish women do not experience a double disadvantage on employment these hypotheses cannot be confirmed for employment: the three forms of capital separately do not explain penalties on employment of Moroccan and Turkish women as compared to native women and their male counterparts. M1 of table 5 has already shown that Moroccan and Turkish women only experience a gender penalty on occupational status as compared to Moroccan and Turkish men. M2 of table 5 shows that the difference in occupational status for Moroccan and Turkish women as compared to their male counterparts and native women is not significant when controlling for human capital. For H10 can thus be concluded that human
23
capital explains the total gender penalty of Moroccan and Turkish women as compared to Moroccan and Turkish men. In other words, when controlling for human capital, Moroccan and Turkish women do not significantly score lower on occupational status than their male counterparts. M3 of table 5 shows that there is still a significant difference on occupational status for Moroccan and Turkish women as compared to Moroccan and Turkish men when controlling for social capital. The B-coefficient on the interaction variables is 7.7 (p-value < 0.01) for female*Moroccan and 7.82 (p-value < 0.01) for female*Turkish. The disadvantage of Moroccan women as compared to Moroccan men is now 5.23 points (-12.93+7.7=-5.23), this is 32.3% lower than the 7.72 points in M1. The disadvantage of Turkish women as compared to Turkish men is now 3.18 points (-11+7.82=-3.18), this is 50.3% lower than the 6.4 points in M1. For H11 can thus be concluded that social capital does partly explain the gender penalty of Moroccan and Turkish women as compared to their male counterparts but that a part remains unexplained. M4 of table 5 shows that there is still a significant difference on occupational status for Moroccan and Turkish women as compared to their male
counterparts when controlling for cultural capital. The B-coefficient on the interaction variables is 7.22 (p-value < 0.01) for female*Moroccan and 6.81 (p-value < 0.01) for female*Turkish. The disadvantage of Moroccan women as compared to Moroccan men is now 4.09 points (-11.31+7.22=-4.09), this is 47% lower than the 7.72 points in M1. The disadvantage of Turkish women as compared to Turkish men is now 3.87 (-11.31+6.81=-3.87), this is 39.5% lower than the 6.4 points in M1. For H12 can thus be concluded that cultural capital does partly explain the gender penalty of Moroccan and Turkish women as compared to their male counterparts but that a part remains unexplained. These results also show that human capital explains most of the differences of Moroccan and Turkish women as compared to their male counterparts and native women. This is also evident when looking at the explained variance of the models. This is 32% when controlling for human capital, 9% when controlling for social capital, and 15% when controlling for cultural capital.
M5 includes ethnicity, gender and all forms of capital in the regression. The hypothesis regarding M5 of table 3 and 5 is that the penalties experienced by Turkish and Moroccan women at employment and occupational status are partly explained by differences in their human-, social-, and cultural capital together (H13). Again, since Moroccan and Turkish women do not experience a double disadvantage on employment this hypothesis cannot be confirmed for employment. M5 of table 5 shows that the difference in occupational status of Moroccan and Turkish women as compared to native women and their male
24
capital. The B-coefficient on occupational status of female*Turkish is 4.11 (p-value < 0.05). This makes the disadvantage as compared to Turkish men an advantage: -2.31+4.11=1.8. For H13 can thus be concluded that the three forms of capital do explain the total gender penalty of Moroccan and Turkish women as compared to their male counterparts. Table 5 also shows that the model that includes the three forms of capital together has the largest explained variance. However, this is only 2% higher than when only including human capital in M2: 34% as compared to 32%. Thus, adding social- and cultural capital does not influence the explained variance much.
Conclusions and discussion
A lot of research has been done on ethnic and gender penalties in the labor market. This research has contributed to this important and interesting literature by examining if
differences in different resources that are assumed to be of positive influence on labor market outcomes on micro-level can explain these penalties. The aim of this research was to answer the following research questions: (1) can differences in human capital, social capital and cultural capital explain ethnic penalties in employment and occupational status of first generation immigrants from Turkey and Morocco as compared to Dutch natives in the Netherlands? And (2) can differences in human capital, social capital and cultural capital explain both ethnic and gender penalties in employment and occupational status of first generation Turkish and Moroccan women? These questions have been addressed by running four multivariate regressions on cross-sectional data from NELLS (de Graaf, Kalmijn, Kraaykamp & Monden, 2010).
Firstly, as hypothesized, the results show that immigrants with a higher educational level, a higher frequency of contact with natives in unions or clubs and a higher frequency of attending the theater are more likely to be employed than immigrants with a lower educational level, a lower frequency of contact with natives in unions or clubs and a lower frequency of attending the theater. However, the other resources that were presumed to be valuable for the likelihood to be employed showed to be not significant. When controlling for all forms of capital together, the frequency of attending an art museum even turns out to be of negative influence on the likelihood to be employed and only educational level remains of positive influence. An explanation for this could be that this research only looks at the active labor force participation and thus only measures whether one is employed or involuntarily
unemployed. This result indicates that employers hire their employees based on their level of education and not on the other examined resources. This result is in line with human capital
25
theory: investments in education are of positive influence on one’s productivity and efficiency (Becker, 1975) and employers rationally hire employees based on their human capital (Gracia, Vazquez & van de Werfhorst, 2014). However, human capital theory also argues that host country human capital is more valuable than country of origin human capital (van Tubergen & van de Werfhorst, 2007). The discrepancy of the results with this argument could be explained by the fact that there is not differentiated between different countries when one has no Dutch diploma. When one has a diploma that is not Dutch it is unknown where this diploma was obtained. Hence, it does not necessarily mean that this diploma is not
transferable to the Dutch labor market or of less quality than a Dutch diploma. However, there cannot be examined where this diploma is obtained so this remains a question.
Secondly, as hypothesized, the results show that immigrants who have a higher educational level, a better understanding of the Dutch language, more frequent contact with Dutch natives in unions or clubs, a Dutch friend, and more frequently attend to all of the high-status activities score higher on occupational high-status than immigrants who have a lower
educational level, a worse understanding of the Dutch language, less frequent contact with Dutch natives in unions or clubs, have no Dutch friend, and less frequently attend to all of the high-status activities. Not in line with the hypotheses is that having a Dutch diploma does not significantly influence occupational status. This could be because of the same problem as stated for employment: it is unclear where one obtained a diploma when it is not in the Netherlands. Another surprising result is that having more frequent contact with Dutch neighbors is of negative influence on occupational status. A possible explanation has to do with reverse causality. Maybe people with a lower occupational status are more likely to live in areas where there is close contact with neighbors whereas people with a higher
occupational status are more likely to live in areas where there is less close contact with neighbors. When controlling for all forms of capital together it turns out that human capital in the forms of educational level and language proficiency are the most of influence on one’s occupational status. Second is cultural capital, and social capital is of the least influence.
Thirdly, as hypothesized, first generation immigrants from Turkey and Morocco experience ethnic penalties on employment and occupational status. Interesting is that Turks experience a larger ethnic penalty on employment than Moroccans and Moroccans experience a larger ethnic penalty on occupational status. Differences in human-, social- and cultural capital separately can partly explain the ethnic penalties of Moroccans and Turks. Differences in human capital even explain the total ethnic penalty of Turks on occupational status. The ethnic penalties on employment and occupational status are the most reduced by human
26
capital, the second most by cultural capital and the least by social capital. Educational level is the most important for explaining ethnic penalties in employment while language proficiency is the most important for explaining ethnic penalties in occupational status. However, these forms of human capital alone can only explain the total ethnic penalty of Turks on
occupational status. For Moroccans the ethnic penalty is only partly explained. When controlling for all forms of capital together the total ethnic penalty of both Moroccans and Turks on occupational status is explained so social-, and cultural capital are also of influence in explaining the ethnic penalty of Moroccans. However, the three forms of capital do still not explain the total ethnic penalties of Moroccans and Turks on employment. The ethnic
penalties are reduced by respectively 33.8% and 29.8%. Hence, a large part of the ethnic penalties of Moroccans and Turks on employment remains unexplained. In other words, the same resources do not guarantee the same outcomes for Moroccans and Turks as for Dutch natives. This was also expected since there are various sources for ethnic penalties. One source can be discrimination but also the way of job searching of immigrants can be less efficient or intense than that of natives (Kalter & Kogan, 2006).
Fourthly, not in line with the hypotheses, the results show that Moroccan and Turkish women do not experience a double disadvantage on employment and occupational status. On occupational status they do experience a gender penalty as compared to their male
counterparts but no ethnic penalty as compared to native women. They do even score higher on occupational status than native women. On employment they do even not experience one of the penalties. This is surprising but the result on employment can possibly be explained by the fact that this research does not take the choice to work into consideration. Whereas previous research found that women from Morocco and Turkey with more traditional gender role attitudes than native women excluded themselves from the labor market (Khoudja & Fleischmann, 2015), this research finds that Moroccan and Turkish women do not have more trouble to find a job than native women and Moroccan and Turkish men. The gender penalty that Moroccan and Turkish women experience on occupational status is explained in total when controlling for human capital. Social- and cultural capital can only partly explain the gender penalty of Moroccan and Turkish women as compared to their male counterparts. Of these two forms of capital, social capital explains the largest part of the gender penalty for Turkish women and cultural capital explains the largest part of the gender penalty for
Moroccan women. The result that Moroccan and Turkish women do not experience an ethnic penalty on occupational status as compared to native women is hard to explain since
