The gender wage gap in job sectors in the Netherlands

(1)

1

The gender wage gap in job sectors in the Netherlands

This thesis explores the effects of job sectors on the gender wage gap in the Netherlands. A regression equation based on the Mincer equation is used to study the research question. The data come from a questionnaire called Het Nationaal Salarisonderzoek 2017. While a lot of studies investigate the gender wage gap, many do not focus on job sectors. The job sectors that are being used in this research are agriculture (1), industry (2), construction industry (3), trading, catering & repair (4), transport (5), business services (6), care and welfare (7), other services (8), government (9) and education (10). With a significance level of 5%, only the sectors for business services and other services seem to have a significant effect on the gender wage gap.

Name: W. Schilder

Student number: 10758968 Thesis supervisor: S. Gaddam

BSc-programme: Economie & Bedrijfskunde Specialization: Finance & Organization Date: January 2018

(2)

2

Statement of originality

This document is written by Student Wessel Schilder who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

1. Introduction

The inequality between men and women has always been a popular research topic. Lately the difference in income between men and women, the so called gender wage gap, has been discussed a lot in the Netherlands. These discussions started for instance with the research Business University Nyenrode published on the 15th _{of November 2017 (Van Muijen & Melse, 2017).}

The survey used in this research is called Het Nationaal Salarisonderzoek 2017 and is carried out by Intermediair in cooperation with Business University Nyenrode. It’s an online survey and uses all kinds of personal and work related questions. Salaried employees could respond to this survey between April and July of 2017. Approximately 70,000 employees responded to the survey.

The research concludes that women still earn less than men. The researchers find that women younger than 36 earn on average 4% less than men (Van Muijen & Melse, 2017). Women older than 36 even earn 8% less than men, on average (Van Muijen & Melse, 2017).

The main reason for this difference in wages might be job satisfaction. One of the findings in the research is that over 55% of the women think that their wages are justifiable, while only 50% of the men think their wages are justifiable. This might be surprising because men earn more than women, but yet they are less satisfied. There are two possible explanations for this problem. The first explanation is that men find the terms of employment the most important aspect of work, while women find the balance between work and private life more important (Nyenrode Business Universiteit, 2017). The second reason is educational attainment. Highly educated employees are often less satisfied with their wages than lower educated employees (Nyenrode Business Universiteit, 2017). Because men are less satisfied about their wages, they negotiate more often about their wages. Women negotiate less often and this creates a wage gap between men and women.

Since there isn’t much research done about the gender wage gap in the Netherlands, this paper will include own research. As noted above the gender wage gap is a popular research question, but most research doesn’t focus on the Netherlands and a lot of academic literature that does, is quite old. This paper contributes to the existing work by using data from different job sectors in the Netherlands to find out the gender wage gap in each sector. It is interesting to look into job sectors specifically as existing research on gender wage gap usually does not focus on them. The research question for this paper will be: “what is the effect of job sectors on the gender wage gap in the Netherlands?”.

The aim of this research is to find out which sector has the largest gender wage gap. The results can drive future research to look into the factors causing this difference. This knowledge might help policy makers in designing better solutions to reduce or close the gender wage gap thereby leading to more equality and less discrimination.

The data needed to answer the research question comes from the Arbeidsaanbodpanel. The Arbeidsaanbodpanel is a questionnaire from Sociaal en Cultureel Planbureau (SCP). This

questionnaire is used to research the labour supply in the Netherlands. The respondents are the Dutch population from 15-66 years old. It started in 1985 and the results used in this paper are from 2012. The data of 2012 uses a total of 4837 respondents. This data will be analysed using STATA and Mincer’s equation will be used to carry out the research.

The rest of the paper is divided into the following sections. Section 2 consists of all relevant literature and background information. Section 3 discusses the model and the data used. Section 4 outlines the results. Section 5 concludes and gives possible suggestions for further research. The last section will consist of possible suggestions to improve this research or suggestions for further research.

(5)

5

2. Literature review

The first part of this section discusses the relevant literature on the gender wage gap. The second part consists of background information and data about gender wage gap in the Netherlands. The final part provides the hypotheses of this research and concludes with a brief summary.

2.1 Main findings

The differences in the gender wage gaps has two main explanations (Blau & Kahn, 1992). The first explanation is gender-specific factors. When analysing the gender wage gap, economists often focus on gender-specific factors (Blau & Kahn, 1994). Research in these gender-specific factors used to suggest that women are less skilled than men, on average. This meant that women were often working in lower-paying sectors and professions (Blau & Kahn, 1994). This resulted in the wage structure having a big impact on the gender wage gap. Wage structure is the second explanation. Wage structure is the differences in prices set on specific labour market skills (Blau & Kahn, 1992). If in two countries women have lower skill-levels than men, but the difference in skill is the same for both countries, the country with the highest return to skill will have the largest gender wage gap. The skill prices depend on a few things (Blau & Kahn, 1994). First of all the skill prices are influenced by relative supplies of these skills. Secondly it’s affected by technology, because high-tech companies pay a lot for highly trained employees. The last and most important factor is wage-setting institutions. The main finding is that the U.S. gender wage gap is larger than the other countries because of the high level of wage inequality. The high returns to skill put a large penalty on low levels of labour market skill (Blau & Kahn, 1992). In the U.S. in the 1970s and 1980s the return to experience increased. Since men were on average more experienced than women, the gender wage gap rose in this period, even when holding all other factors constant (Blau & Kahn, 1994). The same could happen when returns to employment in male-dominated industries and professions would increase.

Blau and Kahn (2017) found that between 1980 and 2010 the gender wage gap substantially reduced. This was the case for both the U.S. and other industrialised countries. The main reasons for this were the improvements of women in education, experience and occupational representation. After 2010, these variables together explained little of the total gender wage gap. Nowadays women exceed men in educational attainment (in the U.S.) and reduced the difference in experience compared to men. From this point the gender-specific factors in industry and profession are the main reasons for explaining the gender wage gap. However, this doesn’t mean that men and women appear to be seen as perfect substitutes (Blau & Kahn, 1994). This is also suggested by the data shown in section 2.2, where you can see that some job sectors are clearly male- or female-dominated.

Black et al. (2008) used black, Hispanic, Asian and non-Hispanic college-educated women in the U.S. to research the gender wage gap. While Blau and Kahn (2017) found that educational attainment and work experience explained very little of the total gender wage gap, Black et al. (2008) found that the raw log wage gap for well educated-women is approximately 30%. Half of this gap is caused by background variables like age, educational attainment and work experience.

Blau and Kahn (2017) also found that by 2010 the gender wage gap was larger for higher skilled females than lower skilled females. Short hours and work interruptions are important in explaining the gender wage gap in highly skilled professions. This suggests that the labour market for executives and highly skilled employees favoured men. The gap in the top of the wage distribution decreases slower than the middle and bottom distribution.

Although almost all literature concludes that women earn less than men, Gayle et al. (2012) find the opposite. Using panel data on executives and controlling for executive rank and background, Gayle et al. (2012) find that women earn a higher compensation than men. This is due to the fact that women executives have a higher pay to performance than men. Women are also promoted more quickly and women are more likely of becoming a CEO than men. Another finding is that women exit

(6)

6 a firm at a higher rate than men. This is the biggest reason for the differences in average and total career compensation of males and females. If someone quits his or her job, this person usually stops working for a certain amount of time. Without work, one doesn’t get paid. The gap doesn’t exist because of unequal pay, but because of the high exit rate of women.

There is also research available about the gender wage gap which is focused on the Netherlands. Albrecht et al. (2004) analysed the gender wage gap across the log wage distributions for males and females who work full-time. The researchers found an average log wage gap of approximately 20%. When comparing the log wage distributions a strong glass ceiling effect was found. This is the case because the further they move up in the distribution, the larger the increase in the gap. The glass ceiling effect “implies that gender (or other) disadvantages are stronger at the top of the hierarchy than at lower levels and that these disadvantages become worse later in a person's career” (Cotter, Hermsen, Ovadia, & Vanneman, 2001). Most Dutch women work part-time. The researchers find that if most women had worked full-time in the Netherlands, the gender wage gap would have been larger than 20%. Around 25% of the gender wage gap is due to differences between males and females in labour market characteristics. The rest can be attributed to differences in rewards for these characteristics.

Research from the U.S. shows that fedominated sectors receive lower wages than male-dominated occupations, on average (De Ruijter, Van Doorne-Huiskes, & Schippers, 2003). Using multi-level modelling techniques, De Ruijter et al. (2003) analysed the occupational gender wage gap in the Dutch labour market. They found that both males and females earn lower wages in female-dominated occupations. This shows the importance of gender on the labour market, because female labour in the Netherlands is undervalued. The main explanation for this gap seems to be the difference in required responsibility. The gap is even larger when a high level of education, skill and responsibility is required, for both male- and female-dominated occupations.

Bakker et al. (1999) also researched the gender wage gap in the Netherlands. Bakker et al. analysed the gender wage gap based on a panel survey. The data included information about wages, education, occupation, human capital, family background and labour force interruptions. They found a log wage difference of 0.363. 0.190 was due to occupational differences. Educational attainment and work experience also played an important role in explaining the gap. Family background only seemed to be important for females.

2.2 Traditional factors affecting the gender wage gap

To understand the gender wage gap a bit better, we need to understand the factors affecting the gender wage gap. Blau and Kahn (2017) made a list of traditional factors affecting the gender wage gap. These factors are commonly used when trying to explain the gender wage gap.

The first factor in the list is the labour force participation. Labour force participation is an important factor because it’s directly linked to supply and demand and thus has a big influence on wages. When the demand for men is high, while the demand for women is low, the wages for men will be higher than the wages for women. This results in a gender wage gap. Supply could also result in a gap. When the supply of men is low, while the supply of women is high, the wages of men are relative high.

The second factor is sample selection bias. Sample selection bias is “the bias in an estimator of a regression coefficient that arises when a selection process influences the availability of data and that process is related to the depended variable. This bias induces correlation between one or more regressors and the regression error” (Stock & Watson, 2012). So when there is sample selection bias the outcome might differ from the “real” outcome, for instance because the sample is too small and doesn’t represent the population good enough.

(7)

7 Education and mathematics test scores are also commonly used to research the gender wage gap. Women used to have a lower educational attainment than men. These days men and women are pretty much equal in terms of educational attainment. Since wages normally depend a lot on educational attainment, it is was very important factor affecting the gender wage gap. As stated above, gender-specific factors like educational attainment and work experience explain very little of the gender wage gap nowadays.

The fourth factor Blau and Kahn mention are labour-force experience and work hours. Traditionally women work less and shorter hours than men. Women work a lot more than, for instance, thirty years ago, but they still work less and shorter hours. The fact that women work less hours than men is shown in graph 2. Women in the Netherlands generally work part-time, while men have full-time jobs. Because work experience is an important factor affecting wages, it is often used in explaining the gender wage gap.

Although there isn’t much recent research done about it, gender differences in formal training is another factor. Men tend to get more training at work than women. This leads to more experience and a favourable position for men. This in turn leads to better jobs and higher wages for men, and thus in a gap.

The next factor is the impact of the gender division of labour and motherhood. Usually, women take care of the children and the house. This leads to an undesirable outcome in the labour market for women. This is the case because, again, women work less and shorter hours and this results in less work experience. This phenomena is also known as the motherhood wage penalty. The motherhood wage penalty is a negative correlation between children and the wage of women (Blau & Kahn, 2017).

Occupations, industries and firms are the seventh mentioned factors on the list. Blau & Kahn mention that these are one of the most important factors of today’s gender wage gap. While gender-specific factors in the U.S became less important, the gender wage gap accounted by factors like occupation and industry rose from 27% in 1980 to 49% in 2010. This happened while the total gender wage gap decreased in the same period. Better jobs for women helped closing the gender wage gap in this period, but a big part of this effect was outweighed by worse return to occupations. The dimension of firms and the female representation across hierarchies in occupations are also affecting the gender wage gap.

The last factor mentioned on the list is labour market discrimination. Since not all of the gender wage gap can be explained by gender-specific differences, discrimination might offer an explanation. Sadly even today discrimination still exists, so it is also a factor affecting the gender wage gap.

2.3 Non-cognitive skills

Psychological attributes or non-cognitive skills are part of the newer explanations for different outcomes in gender (Blau & Kahn, 2017). Today, labour economists are more interested in non-cognitive skills. These non-non-cognitive skills are used to research the effect on labour market outcomes and behaviour (Blau & Kahn, 2017). The most important factor of non-cognitive skills becoming more important is the decreasing effect of explaining gender differences by traditional economic variables (Blau & Kahn, 2017). The traditional economic variables often leave a big part of gender differences unexplained. This led to researchers looking for new possible variables that could explain the unexplained gap.

Negotiation is a non-cognitive skill which seems to have a big effect on the gender wage gap. Multiple researches have found that women are less likely to negotiate than men. The main reason for this is that women are more afraid of what other people think of them. When women negotiate they often feel like they are being too intrusive. The fact that men negotiate more often than women directly affects the differences in wages for men and women.

(8)

8 Croson and Gneezy (2009) found that women are more risk averse than men. Women being more risk averse could lead to relative lower wages than men. This is because risk aversion might have an effect on job performance in certain professions (Blau & Kahn, 2017). Women are also less likely to avoid jobs with high variable pay, partly due to being more risk averse than men (Blau & Kahn, 2017). According to laboratory experiments, men are also more competitive than women (Blau & Kahn, 2017). During an experiment where men and women had to choose between a non-competitive compensation scheme or a competitive compensation scheme, 73% of the men chose for the competitive compensation scheme, while only 35% of the women chose for this scheme. Even the men with a low performance chose the competitive scheme more often than women with a high performance. Women being less competitive could result in a gender wage gap, because it leads to a disadvantage for women in the labour market. Being less competitive could result in a relative lower pay and could lead to women avoiding certain jobs or business settings.

Norms and gender identity also seems to affect the gender wage gap. Men are often the main wage earners. If the woman outearns the man, marriage often ends up in a divorce (Blau & Kahn, 2017). This might result in a wife with a higher potential income than her husband, not working or earning less than predicted. This happens to keep the traditional gender roles intact. So these norms and values affect the gender wage gap, because of the traditional gender roles still playing a part in today’s society.

Even though these effects seem to have an effect, there are still a few problems. First of all, at this point we can’t tell yet if the differences in non-cognitive skills are part of nature or nurture (Blau & Kahn, 2017). Secondly the rewards for different non-cognitive skills are different for men and women. While men could be rewarded for a certain skill, women could be penalized for this same skill (Blau & Kahn, 2017). Since most research about non-cognitive skills is quite new, much of the evidence comes from laboratory experiments. This comes with problems of generalizing these results (Blau & Kahn, 2017). Last of all, even though women seem to be less competitive, this doesn’t have to be a problem. If it’s known that women are less competitive, women could get training in becoming more competitive (Blau & Kahn, 2017). This could solve the problem.

2.4 Trends in the Netherlands

Graph 1: Overview of gender wage gaps, source: OECD 2014

This table from the OECD (2014) shows the gender wage gaps of different countries, including the Netherlands. The gender wage gap is in this case defined as the difference between male and female median wages, divided by the median wages of males. As one can see the Netherlands scores above

(9)

9 average in this graph. Considering this table and the earlier mentioned literature about the gender wage gap in the Netherlands, it seems like the gender wage gap in the Netherlands does exist.

Sector Amount of males

x1000 Amount of females x1000 Percentage females Agriculture 176 81 31.52% Industry 575 174 32.23% Construction industry 453 59 11.52%

Trading, catering & repair

1,089 932 46.12%

Transport 310 99 24.21%

Business services 136 105 43.57%

Care and welfare 307 1,256 80.36%

Other services 623 823 56.92%

Government 306 206 40.23%

Education 220 367 62.52%

Table 1: Labourvolume, source: Centraal Bureau voor de Statistiek (2017)

In the table above the amount of males and females in each job sector is shown. Sectors with a gender rate higher than 60% will be considered male- or female-dominated. In this case there are only two sectors female-dominated. Only education and care and welfare are considered female-dominated. There are more male-dominated sectors. Agriculture, industry, construction industry and transport are all considered male-dominated sectors. This information helps to interpret the results from section 4 better.

Graph 2: Average working hours per gender per week, source: Centraal Bureau voor de Statistiek 2015 As the graph shows, women in the Netherlands make a lot less working hours on average. Women in the Netherlands start working part-time from a young age (SCP, 2017). 62% of the women between 18 and 25 work time, while only 28% of the men work time (SCP, 2017). Women working part-time often results to being less economically independent than men. 66% of the women between 30 and 34 are economically independent and 82% of the men are economically independent (SCP, 2017). This doesn’t seem to be a result of the availability of jobs, because after graduating men and women

0 5 10 15 20 25 30 35 40 45 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Average working hours per gender per week

(10)

10 are equally likely of finding a job (SCP, 2017). The Netherlands has the biggest difference in working hours of men and women of Europe. In the Netherlands the average difference is eight hours, while the average difference in the rest of Europe is four hours (SCP, 2017). You could conclude that women, on average, still have less work experience than men. Women work on average less than men. While men work on average between 35-40 hours per week, women only work around 25 hours per week. This results in an experience gap between men and women.

Graph 3: Highly educated males and females in the Netherlands, source: Centraal Bureau voor de Statistiek 2017

The graph above shows the amount of degrees for highly educated men and women in the Netherlands. CBS states that someone is highly educated when this person has at least a “hbo-bachelor” (Centraal Bureau voor de Statistiek, 2008). The graph shows that there are still more highly educated men than women, but if growth rates stay like this, than in a few years there will be more highly educated women than men. This shows that Blau and Kahn (2017) were probably right in stating that educational attainment explains little of the gender wage gap nowadays.

0 500 1000 1500 2000 2500 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Amo u n t o f d eg ree s x1 ,0 0 0 Years

Highly Educated Males And Females In The Netherlands

(11)

11 Graph 4: Average number of children per female in the Netherlands, source: Centraal Bureau voor de Statistiek 2017

The graph shows the amount of birth since 1950 till 2016. The amount of births decreased in this period with 46.30%. A big decrease in average children happened between 1969 and 1973.

As stated earlier, having children results in an effect called the motherhood wage penalty (Blau & Kahn, 2017). This effect states that having children has a negative correlation with income. So more children lead to a bigger penalty on income. This usually counts for women. Most women take care of the children and because of this get less working experience, since they stop working or start working less hours. One of the reasons the gender wage gap in the Netherlands should have improved is because of the decrease in average children between 1950 and 2016.

Graph 5: Labour force participation in the Netherlands, source: Centraal Bureau voor de Statistiek 2015 As mentioned in the list of Blau and Kahn (2017) an important factor on the gender wage gap is labour force participation. It’s an important factor because wages depend a lot on supply and demand and

0 0,5 1 1,5 2 2,5 3 3,5 19 50 1 9 53 1 9 56 1 9 59 1 9 62 1 9 65 1 9 68 1 9 71 1 9 74 1 9 77 19 80 19 83 1 9 86 1 9 89 1 9 92 1 9 95 1 9 98 2 0 01 2 0 04 2 0 07 2 0 10 20 13 20 16 Amo u n t Of C h ild ren Years

Average Number Of Children Per Female

Children 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Amo u n t x1 ,0 0 0 Year

Labour Force Participation

(12)

12 the labour force participation is directly linked to this. As shown in the graph the labour participation of men didn’t increase much. Their labour participation went up with approximately 3%. The participation of women increased more. In the period of 1996-2014 the women labour participation went up by approximately 37%. The increase of the participation of women has different reasons. The main reason seems to be the increased demand for highly-educated employees. With today’s technology, intellectuality becomes more and more important and this results in an increasing demand for highly-educated women. Also, since 2000 there is a new law which makes it possible to work part-time. Since 2000 the right exists to work part-part-time. Before, one often worked full-time or didn’t work. The right to work part-time also helped the labour participation of women grow.

2.5 Hypotheses

Two hypotheses will be used to answer the research question. Based on the literature review above the first hypothesis will be that the gender wage gap in the Netherlands exists, and thus the effect of male is significant. The second hypothesis will be that job sector does have an impact on the gender wage gap. The expectation is that the effect of male-dominated sectors is larger than the effect of female-dominated sectors. To summarize the hypotheses are as follows:

H0: male = 0 H1: male ≠ 0 H0: Zi = 0 H1: Zi ≠ 0

In this case male is a dummy variable, where 1 means the respondent is a male and 0 indicates a female. Zi indicates the job sectors that are being used in the model. The job sectors are listed in section 3.2.

2.6 Summary

There are a lot of factors affecting the gender wage gap. The estimated gender wage gap seems to be very dependent on the used data. Most papers conclude that the gender wage gap exists, but the outcomes are never the same. While Blau and Kahn (2017) find that educational attainment has almost no effect on the gender wage gap, Black et al. (2008) find almost the complete opposite. Gayle et al. (2012) are one of the few that conclude that women earn more than men, instead of the other way around. Gender-specific factors like educational attainment and work experience used to be one of the main explanatory variables of the gender wage gap. Nowadays wage structure and gender-specific factors in industry and profession seem to be the most important factors affecting the gender wage gap. Research in non-cognitive skills also became important in explaining the gender wage gap. Since the traditional economic variables often leave an unexplained gap, researchers moved to other possible factors affecting the gender wage gap. Non-cognitive skills like negotiation seem to play an important role in explaining the gender wage gap.

According to the research the gender wage gap still exists. By looking at the trends in the Netherlands and focusing on gender-specific factors, it shouldn’t take long for women to earn the same as men. There are nearly as many highly-educated women as men and the average number of children is decreasing. These factor should both result in an decreasing gender wage gap. The main reason for the existing gender wage gap seems to be the part-time jobs women have and the wage structure.

This paper will focus on the effect of ten different job sectors on the gender wage gap in the Netherlands. The effect will be shown by running an ordinarily least squares (OLS) regression in Stata. Section 3 presents a detailed analysis of the research method.

(13)

13

3. Research method

As noted in the previous section, an OLS regression will be run to show the effect of job sectors on the gender wage gap in the Netherlands. An OLS regression line uses estimators that minimize the sum of squared residuals (Stock & Watson, 2012). These estimators replace the population coefficients to estimate the outcome.

The Mincer equation is used to carry out the regressions. The following subsections discuss the Mincer equation in detail, the data being used and the model.

3.1 Mincer equation

Jacob Mincer published his book Schooling, Experience and Earnings in 1974. His book had a lot of influence on empirical work in labour economics (Lemieux, 2006). The model is used to estimate return to schooling, return to schooling quality and measure the effect of work experience on the gender wage gap (Heckman, Lochner, & Todd, 2003).

Mincer used a model with the dependent variable the natural logarithm of earnings. The independent variables he used are years of education and years of potential labour market experience. Potential labour market experience is defined as age – years of education – 6. The most commonly used Mincer equation is as follows:

log y = log y0 + rS + β1x + β2x2 + εi

In this equation y is earnings and y0 is the amount of earnings of a person without education nor experience. S means years of schooling and X is years of potential labour market experience. r, β1, β2 are the regression coefficients and εi is the error-term.

More than forty years after Schooling, Experience and Earnings a lot of studies still rely on this equation (Lemieux, 2006). Often extra variables are added to the equation, but the three independent variables and the logarithmic specification for earnings are almost always used.

As stated above, the dependent variable is the logarithm of earnings. Logarithms are often used for convenience or fit (Stock & Watson, 2012). In this model the logarithm of earnings is used to give years of schooling the desired effect on earnings (Mincer, 1958). Mincer used a linear specification for years of schooling. This is easy to interpret, but Lemieux (2006) states that there are several reasons why a linear years of schooling specification might be inaccurate. The average logarithm of earnings is more likely to be convex or concave for instance (Lemieux, 2006). The quadratic function of years in potential labour market experience is used “to capture the fact that on-the-job training investments decline over time in a standard lifecycle human capital model” (Lemieux, 2006). It’s also more convenient than estimating, for instance, a complex non-linear specification which might be more fitting with economic theory.

3.2 Data

As stated in the introduction, the needed data comes from a questionnaire called the Arbeidsaanbodpanel. This data has been put in Stata and will be used to run multiple regressions based on the Mincer equation. Not all the data from the dataset is relevant for the model being used in this paper. The main independent variables are years of schooling, which is labelled as LevelOfEducation, labour market experience, male (a dummy variable equal to 1 for males and 0 for females) and interaction variables including male and job sector. The job sectors that will be used are: agriculture (1), industry (2), construction industry (3), trading, catering & repair (4), transport (5), business services (6), care and welfare (7), other services (8), government (9) and education (10).

The data has been sorted out and modified to fit the regression model. First of all, students and people above 65-years old have been dropped. Only salaried employees are considered for the regressions. Respondents with weird or incomplete data have also been dropped. For instance, one

(14)

14 respondent was 58-years old with 51 years of labour market experience. Respondents like this have been dropped. Furthermore, new variables have been created. All used variables are shown in table 2. Since all data was coded, variables like ab032_12 are renamed, in this case to “children”. At the end 2573 respondents were used to run the regression.

3.3 Model

The Mincer equation is used as base model, but to answer the research question more variables should be added to the Mincer equation. Since the effect of job sector on the gender wage gap is researched, the dummy variable male should be added. Also control variables are added to complete the regression model. The following regression model is used:

log y = β0 + rS + β1x + β2x2 + male + control variables

Since, it is difficult to convert wages per month from the dataset into net wages per hour, y in this case will be the net wages per month. Y0 is replaced with β0 and will be estimated by Stata. S indicates the level of education instead of years of schooling. Actual labour market experience instead of potential labour market experience is used, because the data included actual labour market experience. The rest of the used variables are described in 3.1 and 3.2.

The equation quantifies as a log-linear function. This means that when, for instance, level of education goes up by one, y changes by 100% times the regression coefficient (Stock & Watson, 2012).

(15)

15

4. Results

In this section the results from the regressions will be discussed. The first output will discuss the gender wage gap over the whole dataset. For this output all variables will be discussed. The following subsections discuss the output for all the job sectors and how the results affect the hypotheses.

4.1 Regression results

As stated above, the output of the whole dataset will be discussed first. Since all regressions are basically the same in terms of variables, this regression will be analysed extensively. The rest of the regressions will focus on the effect of job sector or results that differ from the results of the first discussed output. The regression model is a log-linear model. When an independent variable changes by one unit, the change in y is associated with 100% times the regression coefficient (Stock & Watson, 2012). To interpret the results as the effect on net wages per month, all regression coefficients will be exponentiated

.

An α of 5% will be used to test significance. A t-value bigger than |1.96| or a p-value < 0.05 indicate that a variable is significant.

The effect of level of education on net wages per month for the first regression output has a very high significance level. Increasing level of education by one leads to an increase of net wages per month of 24.10%. Labour market experience also has a significant effect on the net wages per month. Every extra year of labour market experience results in an increase of net wages per month by 2.45%. Labour market experience squared has a negative effect on the net wages per month -0.03%. The effect is also significant. This means that at a certain point labour market experience squared will overshadow labour market experience. This is probably because at a certain point one is too old to work efficiently and hence their net wages per month will decrease. The estimated effect of male has a significant regression coefficient of .1864237. This means, that a man, on an average earns 20.49% higher net wages per month than that of a woman. This is also the estimated gender wage gap. For each child there is a significant effect of -4.26% on the net wages per month, but every child living at home has an insignificant effect of 1.64%. The effect of a male working full-time obviously has a significant effect on the wages per month. The estimated effect is 41.29%. Working full-time means working more hours and thus a higher wage per month. Marital status has a significant and positive effect on the net wages per month of 3.39%. Country of birth of the respondent, the respondent’s mother and the respondent’s father all have an insignificant effect on the net wages per month, just like the level of education of the respondent’s parents. The estimated effects of job satisfaction and health are both negative. The effect of job satisfaction is even significant with a t-value of -4.92. This is an interesting result, because one would expect these effects to be positive. It might be the case that if someone is satisfied with their job, they don’t negotiate for a better pay. This could result in a negative effect of job satisfaction on the net wages per month. Having a partner results in positive effect of 4.89% on the net wages per month. β0 in this regression output has a coefficient of 5.849. The exponentiated value therefore is 346.961. This means the geometric mean of net wages per month is €346.96. Since these results have been discussed quite extensively, the rest of the regression outputs will mainly focus on the effect of male. All of the regression outputs can be found in the appendix. The first job sector on the list is the agriculture sector. After sorting and modifying the data, the data for the agriculture sector is no longer usable. The agriculture sector only has 16 observations left and this is not enough to run a regression. The agriculture sector won’t be discussed.

Industry is the second job sector for which the results will be shown. In the industry regression output, male has an effect of 2.28% on the net wages per month. This effect is insignificant, but would mean that a man in the industry sector earns 2.28% per month more on average than a woman. Since the industry sector is considered a male-dominated industry, this result is in line with the expectation. However, we would also expect a significant effect.

(16)

16 The third discussed sector is the construction industry. The effect of male on the net wages per month, for this regression, is 4.31%. Just like the industry sector, the construction industry is considered a male-dominated industry. Even though the job sector has a positive effect, it is insignificant. Due to collinearity the variable CountryOfBirthMother has been omitted.

The trading, catering & repair is the fourth discussed sector. In this regression output the effect of male results in a net wages per month of 3.89% higher than women. The effect is again insignificant. This result shows that being a man in the trading, catering & repair sector has a positive effect on net wages per month.

The next sector is the transport sector. The found effect of male on the wages per month for this regression is 11.40%. This variable also has a high insignificance level, even though it’s assumed to be a male-dominated sector.

Business services is the sixth sector. The regression output shows an effect of 24.07% for males. Being a male in the business services sector leads to a higher net wage per month of 24.07% than females. With a t-value of 2.57 this is the first regression with a significant effect for male.

The regression output for care & welfare will be analysed next. The found effect is 11.85%. Even though this sector is assumed to be female-dominated, the effect of male is insignificant. Apparently the effect of being a male in this sector doesn’t affect the wages per month significantly. What is interesting to see in this output is that the variables for labour market experience have high levels of insignificance. Apparently labour market experience doesn’t affect the net wages per month. The effect of male on the log wages per month in the regression output of the other services sector is 38.33%. A male working in this sector earns a net wage per month of 46.71% more than females. With a t-value of 2.64, this effect is significant. This is an interesting result since, according to the data, males earn almost 50% more than females in this sector. Also in this output the variables for labour market experience don’t seem to affect the net wages per month significantly.

The government job sector is the ninth job sector that will be discussed. A male working in this job sector earns 11.74% per month more than females. With a p-value of 0.058 this effect is insignificant but it is close to significance.

The last job sector is the education sector. Being a male in the education sector has an effect of 10.01% on net wages per month. This sector is assumed to be female-dominated, but being a male in this sector doesn’t have a significant effect on the net wages per month.

4.2 The existence of the gender wage gap in the Netherlands

According to the results of the first regression output, the gender wage gap in the Netherland is 20.49%. This is the effect of the dummy variable male on the net wages per month. This means that a man in the Netherlands earn on average 20.49% more than a woman. This effect is significant. This shows that there is a clear gender wage gap in the Netherlands and thus the first hypothesis seems to be true.

4.3 The effect of job sector on the gender wage gap

The expectation was that job sectors would have an effect on the gender wage gap, and that this effect would be significant. This seems only to be partly true. All regressions show a positive effect of being a male in the nine analysed sectors. However, most of these effects are insignificant. Industry, construction industry, trading, catering & repair, transport, care & welfare and education are all insignificant. Government is close to significance but with a p-value of 0.058 and an α of 5%, this effect is also insignificant. Only the sectors for other services and business services show significant results for male.

The results show a gender wage gap for the male-dominated sectors industry, construction industry and transport of 2.28%, 4.31% and 11.40% respectively. The female-dominated sector care &

(17)

17 welfare and education show gender wage gaps of 11.85% and 10.01% respectively. The trading, catering & repair sector shows a gender wage gap of 3.89%. Business services show an effect of male of 24.07%. The other services sector shows a gender wage gap of 46.71% and government shows an effect of 11.74%. Only business services and other services show significant results for male. These results also show the biggest gender wage gap. The expectation was that the job sector would have an effect on the gender wage gap and that the gap in female-dominated sectors would be lower than the gap in male-dominated sectors. Looking at the results this doesn’t seem to be true. While the gender wage gap in female-dominated sectors is around 11%, the gender wage gap in male-dominated sectors is on average 6%. Also, these effects are all insignificant.

4.4 Discussion

Some of the results differ from the expectations. This might has different reasons. First of all, the amount of observations have decreased after going through the data. The sector results might differ from their real values because of this. Agriculture has been omitted for this reason. The other services sector has only sixty male observations left. Even though this result is significant, with sixty observations this result might not show the real outcome. Sample selection bias might be another explanation. The expectation could also be simply wrong. While we expect one thing to happen, the could opposite could be the case. The first two reasons also effect the level of significance.

Other unexpected findings are the effects of job satisfaction and health on the net wages per month. These effects are found to be negative, while you expect these results to be positive. Especially job satisfaction has an important effect, since this effect is also found to be significant. Normally, we would say that if someone has a high job satisfaction, he or she would also have a higher wage per month.

As stated in section 2, there are a lot of factors affecting the gender wage gap. Factors like education, work experience and non-cognitive skills all affect the gender wage gap. Graph 2 shows that women in the Netherlands work on average fewer hours per week than men. Men in the Netherlands work between 35-40 hours per week, on average. Women only work around 25 hours per week. This means that a lot more women work part-time in the Netherlands. The results show a gender wage gap of 17.56%-20.62%. Since the dependent variable is the natural logarithm of net wages per month, worked hours play an important role in the outcome. The main reason for the large gender wage gap is probably the average worked hours. Because women generally work part-time, while men work full-time, women work fewer hours per month. This results in a lower net wage per month. The outcome is probably still because of the differences in gender, but the main reason why the results show a quite big gender wage gap is the difference in average worked hours.

As described in section 3.1 the Mincer equation might also be the reason for weird outcomes. It is useful to use the natural logarithm of net wages per month for convenience or fit. Mincer used a linear specification for years of schooling. Lemieux (2006) states that there are multiple reasons why this might be inaccurate. This specification is more likely to be convex or concave for instance. Also the quadratic specification of potential labour market experience might be wrong. A complex non-linear specification might be more fitting with economic theory (Lemieux, 2006).

In table 3 in the appendix are the results shown for a test for multicollinearity. A VIF-value higher than 10 shows significant signs of multicollinearity. Cases of imperfect multicollinearity, a case which occurs if regressors are highly correlated, do not prevent estimation of the regression (Stock & Watson, 2012). It could however result in imprecisely estimated regression coefficients. Only the variables for labour market experience show significant results for multicollinearity, so multicollinearity doesn’t seem to play a big part in the estimated results.

(18)

18

5. Conclusion

The gender wage gap has been a popular research topic for a long time now. Most researches seem to agree about the factors affecting the gap, but the outcome of a research is never the same. For instance with Black et al (2008) and Blau and Kahn (2017). 15% of their found wage gap came from variables like educational attainment or work experience. Blau and Kahn, however, state that factors like educational and work experience explain little of today’s gender wage gap. All these different outcomes are probably due to the different research methods. A lot of researches use their own data or their own regression equation. This results in different results for every research.

In this paper an OLS regression equation is used, based on the Mincer equation. This equation is used to answer the research question: “what is the effect of job sectors on the gender wage gap in the Netherlands?”. Two hypotheses were used to help give an answer to this question. The first hypothesis is that the gender wage gap in the Netherlands exists. The second hypothesis is that jobs sectors do have a significant effect on the gender wage gap.

According to the results, the first hypothesis can’t be rejected. The first regression output shows a clear gender wage gap 20.49%. These results show the effect of being a male on the net wages per month. Section 2 showed that this could have a variety of reasons. Gender-specific factors like educational attainment and work experience used to be the most important factors, but this has changed of the past few years. Nowadays, for example, non-cognitive skills are used to explain the gender wage gap. Next to these differences in gender characteristics, the main reason for this large gap is probably due to the fact that it is about wages per month. Since females in the Netherlands generally work part-time a large part of the gap could be accounted to this.

The second hypothesis shows different results than expected. Being a male in the researched job sectors seem to affect the net wages per month, but these effects are often insignificant. We also expected the outcomes to differ for male- and female-dominated sectors. The expectation was that both male- and dominated sectors would affect the gender wage gap, but the effect for female-dominated would be smaller. The results show that the gap in female-female-dominated sectors is larger than the gaps in male-dominated sectors. Only the sectors for business services and other services have a significant effect on the gender wage gap. Even though most findings are insignificant, job sector seems to have an effect on the gender wage gap in the Netherlands.

(19)

19

6. Suggestions for further research

The current research can be improved in multiple ways. A larger and different dataset will yield different results which might validate or invalidate the current results. In the current dataset the amount of males in the agriculture and other services sectors are lower than sixty. A larger dataset might resolve this problem. The same regression model could be extended to another country. In this paper, the results show that men in female-dominated sectors earn less than women. Further research could try to explain why this is the case. Instead of using the same regression equation a new research could also estimate the effect by using a more complex regression equation. As stated in part 4.4, the Mincer equation might not be the best equation to test the gender wage gap. A more complex equation might result in better and more in depth results. In this paper the effect of job satisfaction on the net wages per month is negative and significant. This effect could be researched further.

(20)

20

References

Albrecht, J., Van Vuuren, A., & Vroman, S. (2004). Decomposing the Gender Wage Gap in the Netherlands with Sample Selection Adjustments. IZA Discussion Paper, 1-50.

Bakker, B. F., Tijdens, K. G., & Winkels, J. W. (1999). Explaining gender wage differences. Netherlans Official Statistics , 36-41.

Blau, F. D., & Kahn, L. M. (1992). The Gender Earnings Gap: Learning from International Comparisons. The American Economic Review, 533-538.

Blau, F. D., & Kahn, L. M. (1994). Rising Wage Inequality and the U.S. Gender Gap. The American Economic Review, 23-28.

Blau, F. D., & Kahn, L. M. (2017). The Gender Wage Gap: Extent, Trends, and Explanations. Journal of Economic Literature, 789-866.

Centraal Bureau voor de Statistiek. (2008, Januari). Opleidingsniveau. Opgehaald van cbs.nl: https://www.cbs.nl/nl-nl/artikelen/nieuws/2008/16/bijna-evenveel-hoogopgeleide-als-laagopgeleide-nederlanders/opleidingsniveau

Cotter, D. A., Hermsen, J. M., Ovadia, S., & Vanneman, R. (2001). The Glass Ceiling Effect". Social Forces, 655-681.

Croson, R., & Gneezy, U. (2009). Gender Differnces in Preferences. Journal of Economic Literature, 448-474.

De Ruijter, J. M., Van Doorne-Huiskes, A., & Schippers, J. J. (2003). Sizes and Causes of the Occupational Gender Wage-gap in the Netherlands. European Sociological Review, 345-360.

Gayle, G.-L., Golan, L., & Miller, R. A. (2012). Gender Differences in Executive Compensation and Job Mobility. Journal of Labor Economics, 829-871.

Heckman, J. J., Lochner, L. J., & Todd, P. E. (2003). FIFTY YEARS OF MINCER EARNINGS REGRESSIONS. NATIONAL BUREAU OF ECONOMIC RESEARCH, 1-71.

Lemieux, T. (2006). THE ‘‘MINCER EQUATION’’ THIRTY YEARS AFTER SCHOOLING EXPERIENCE, AND EARNINGS. New York: Springer.

Mincer, J. (1958). Investment in Human Capital and Personal Income Distribution. Journal of Political Economy, 281-302.

Nyenrode Business Universiteit. (2017, november 15). Vrouwelijke 35-plussers verdienen nog steeds

8% minder dan mannen. Opgehaald van Nyenrode.nl:

https://www.nyenrode.nl/nieuws/n/vrouwelijke-35-plussers-verdienen-nog-steeds-8-minder-dan-mannen

SCP. (2017, januari 31). Nederlandse vrouwen werken al op jonge leeftijd in deeltijd. Opgehaald van SCP.nl:

https://www.scp.nl/Nieuws/Nederlandse_vrouwen_werken_al_op_jonge_leeftijd_in_deeltij d

Stock, J. H., & Watson, M. W. (2012). Introduction to econometrics. Essex: Pearson.

(21)

21

Appendix

Table 2: Variable descriptions

Name Description

LevelOfEducation Level of education, on a scale from 2-6

WorkExperience Years of labour market experience

WorkExperience2 Years of labour market experience squared

Male Dummy variable for male (1=male, 0=female)

children The amount of children of the respondent

ChildrenLivingAtHome The amount of children of the respondent, living at home

MaleFulltime Interaction variable for male and fulltime employment

MaritalStatus Marital status

CountryOfBirth Respondent’s country of birth

CountyOfBirthFather Country of birth respondent’s father

CountryOfBirthMother Country of birth respondent’s mother

JobSatisfaction Job satisfaction, on a scale of 1-4

Health Health, on a scale of 1-5

LevelOfEducationFather Level of education respondent’s father, on a scale from 2-6

LevelOfEducationMother Level of education respondent’s mother, on a scale from 2-6

partner Dummy variable for partner (0=no, 1=yes)

(22)

22

(23)

23

(24)

24

(25)

25

(26)

26

(27)

27

(28)

28

(29)

29

(30)

30

(31)

31

(32)

32

(33)

33

The gender wage gap in job sectors in the Netherlands