Female Disadvantage in the Gender Wage Differential of the Netherlands
To what extent do marital status, ethnicity and job type explain discrimination?
ABSTRACT
The gender wage differential is a persistent aspect of the Dutch labour market. This study examines the gender wage differential across different levels of income by accounting for marital status, ethnicity and job type via field of education. Panel data representative of the Dutch population from the years 2011-2014 is used to study the wage differential. The results of the Oaxaca-Ransom decomposition find an overall 14% wage gap for the Netherlands with female disadvantage arising from marital status and female advantage arising from field of education. The Melly decomposition indicates an increasing level of discrimination for lower levels of income.
Shreya Ashu Goel[1]
(368890)
Erasmus University Rotterdam
Erasmus School of Economics
Bachelor Thesis: Labour Economics
Supervisor: Dr. Bas Karreman
July 2015
Goel 19
TABLE OF CONTENTS
Introduction2
Theoretical Framework4
Existence of a persistent wage gap5
Measuring wage differentials6
Literature Review8
Independent Variables 8
Level of income 8
Race/ethnicity 9
Civil Status 9
Control Variables10
Job nature 10
Other controls 11
Data13
Methodology17
Results19
Conclusion24
Bibliography26
INTRODUCTION
The gender wage differential is a persistent aspect of the labour market and has been an important subject of research as well as politics and legislature. Even in the European labour market, on average women earn 16% less than men even though women represent 60% of the undergraduate population (European Commission, 2014). However, under the European Union (EU) law, wage discrimination at the workplace is prohibited. The EU has laid out several directives in order to reduce the gender wage differential as well as the gender gap in employment rates. Despite several directives and wage discrimination laws, the persistence of the wage gap requires a further breakdown of the wage gap to explain the causes and initiate dialogue for improvement (Christofides, Polycarpou, & Vrachimis, 2013).
Over the past 40 years, the role of women has shifted dramatically in the European labour market; particularly in the Netherlands. Female labour participation rate has risen from 30% to 70%. Moreover, 44% of Dutch women complete tertiary education whereas only 38% of the Dutch men do so. Albeit this shift in the landscape, a 2014 European commission report indicated that the gender wage gap in the Netherlands is 16.9%, higher than the EU average (Statistics Netherlands, 2014).
Mincer’s (1974) human capital model proposes a simple explanation for wage gaps. The model uses education and training as indicators of input which are to fully reflect on the output indicated by wages. Although the model has provided sound evidence for the wage gap between individuals with primary, secondary and tertiary education, the model has shown to be highly idealistic and is unable to explain several irregularities found in the labour market, especially the gender pay gap.
There exists a general consensus that the gender wage gap is a complex matter and is a result of numerous interrelated observable and unobservable issues. The debate regarding this issue not only is centred on the causes of this gap but also around the fairness of these pay discrepancies. Several factors unrelated to human capital investments have been shown to widen the pay gap. Women and men typically tend to work in different sectors and have different jobs. The female dominated sectors have shown to have lower wages than the male dominated sectors. Furthermore, on average, women earn less than men in female dominated sectors indicated the possibility of the undervaluation of women’s skills. Family responsibilities and traditional gender roles have also shown to widen the pay gap by serving as a nudge towards having a part-time job (European Commission, 2014).
The aforementioned statistics about the Netherlands question the idea of human capital differences accounting for the gender wage gap. Hence, other discriminatory factors such as job nature (Weinberger, 1998; Barker, 1993; Albrecht, Vuuren, & Vroman, 2009) , race and ethnicity (Green & Ferber, 2005; Hofer, Titelbach, & Winter-Ebmer, 2014; Weinberger, 1998), and marital status in combination with sexual orientation (Barker, 1993; Elmslie & Tebaldi, 2014; Green & Ferber, 2005) must be accounted for in an attempt to explain the portion the wage gap not explained by the human capital theory.
Hence, this paper attempts to answer the question to what extent does job type, race, ethnicity, marital status explain the overall gender wage differential across various income levels in the Netherlands? Furthermore, the paper specifically aims to answer the following sub-questions:
· Which of the addressed discriminatory variables lean towards male advantage portion and the female disadvantage portion of the overall wage differential?
· What is the relation between the size of the wage differential and the level of income within the Netherlands?
o What is the relation between the levels of discrimination based on the aforementioned variables and the level of income within the Netherlands?
The remainder of this paper is structured as follows: section 2 discusses the two main neo-classical theories that explain gender differential followed by a discussion of decomposition methods. Section 3 reviews several empirical papers and their findings about the gender wage gap. The review includes a discussion of the nature of the impact of certain variables on wages and forms hypotheses. Section 4 describes the data used for the Netherland and section 5 lays out the model and the used methodologies. Finally, section 6 presents the results for the hypotheses followed by the construction of the overall wage differential from the results of the hypotheses. Lastly, a summary and policy recommendations are presented in section 7.
THEORETICAL FRAMEWORK
i. Existence of a persistent wage gap
As pointed out by Arrow (1973), standard economic theory predicts that all discrepancies in wages can be explained by differences in productivity. Furthermore, the theory indicates a positive relation between the amount of human capital investment and productivity of an individual. The type of human capital investment is indicative of the category of skills that increase. Hence, discrimination can be defined as the wage discrepancies arising from characteristics other than productivity and human capital. Even though the human capital and productivity approach explain a large part of the wage gap, almost 38% percent of the gender wage gap is unexplained (Blau, Ferber, & Winkler, 2002).[2]
There are two traditional theoretical approaches that provide explanation for the gender wage differentials (Green & Ferber, 2005). The first theory explaining the gender wage differential is Becker’s “taste theory”. Becker defined discrimination in terms of an aversion towards interaction with other races. According to the taste theory, employers may regard employees of another race more expensive than they truly are (Becker, 1957). Although this argument does not hold in a perfectly competitive market, in a market where firms have a certain form of monopoly power, there is a possibility of discrimination. Furthermore, it is also acknowledged that certain employers may be willing to choose employers of one race over another even though it may result in lower profits. The following theory can be extended to gender discrimination where employers are willing to accept lower profits by mainly hiring male employees and regarding female employees more expensive than they really are (Green & Ferber, 2005).
In addition to the taste theory, Becker also hypothesizes that married women earn less than unmarried women and married men albeit no differences in productivity. Given that men and women do not differ in comparative advantages for household activities, the traditional division of responsibility, where women are responsible for household activities such as childcare, the wages of married women will be lower. The energy spent by females on household activities acts as a disincentive for them to invest in human capital (Becker, 1985).
The second traditional theoretical approach is an extension of the human capital model that accounts for heterogeneity in human capital and occupational choices (Polacheck, 1981). Polacheck (1981) presents the problem in the form of maximizing lifetime income in the following form:
maxS δT-H-S Wδ,I K(S,δ)
Equation 1 indicates the maximization problem for lifetime income in terms of retirement age (T), years spent out of the labor force (H), years spent on schooling (S), wage (W), occupational characteristics (δ) and individual characteristics (I). Polachek proposed that these occupational characteristics were positively related to wages and that a high δ would indicate a faster growth in wages. The maximization problem of equation 1 indicates that individuals with a higher H would choose a job with a lower δ.
Polachek extended the aforementioned model to gender differences, indicating that women invest in human capital differently as they expect their participation to be intermittent. Furthermore, occupational characteristics are effected by exogenous variables such as marital status and the age of the youngest child when comparing the occupational choice for men and women. The aforementioned hypothesis is similar to the one provided by Becker (1985).
The aforementioned theories highlight several limitations present in the human capital model. Polachek’s model indicated that accounting for human capital was not sufficient to account for wage differences and introduced the inclusion of other occupational characteristics. Furthermore, Becker’s theories suggest the inclusion of variables such as marital status as well as stresses on the presence of unknown factors leading to a “taste” for discrimination. Hence, the aforementioned theories indicate the need for including several other unobservable variables behind the gender wage differential.
ii. Measuring wage differentials
In order to analyse a persistent wage gap between two groups of people (for example, race or gender), one of the most common approach followed by several researchers is the Oaxaca-Blinder (O-B) decompositions (Elder, Goddeeris, & Haider, 2010). This approach allows for the following bifurcation of the gender wage differential: a portion attributable to differences in qualifications (q) and a portion attributable to discrimination (d). The analysis begins by estimating a simple equation for the compensation of men and women individually:
Wg=βg0+βgiXgi+εg where i=1,….n, g=m,f
Equation 2 expresses the wage of an individual from a particular gender as a function of a set of explanatory variables denoted by the vector Xgi where i indicates the type of variable and g indicates whether the indicator is for male (m) or female (f). Furthermore, βg0 denotes the intercept of the estimation and βgi is the vector of coefficients for the explanatory variables. Following the estimation of individual equations, the decomposition is computed as follows:
Wm-Wftotaldifferential=Xm-Xfβmq+(βm-βf)Xfd
Equation 3 shows the mean difference in the earnings of the two genders. Furthermore, the estimation above uses the male distribution as the reference. Both, Oaxaca (1973) and Blinder (1973) acknowledge that using the female distribution as a references i.e. using βf to estimate q is a viable substitute. However, the presence of this alternative has led to ambiguity when quantifying productivity related differences which arises from the arbitrary nature of choosing the reference group. The ambiguity is caused by different reference groups leading to different values for the wage differential (Elder, Goddeeris, & Haider, 2010).
Given the ambiguity, Oaxaca & Ransom (1994; 1999) suggest the O-B method of using one group’s distribution as the norm as being too extreme of an approach. The suggested alternative is an extension of the simple decomposition method and employs pooled estimators (Neumark, 1988). Apart from estimating individual equations for both the genders, this method also requires the estimation of a general equation for the entire sample. One of the widely used pooled estimator method suggested by Oaxaca & Ransom (1994; 1999) is given as follows:
Wm-Wftotaldifferential=Xm-XfβPq+(βm-βP)Xmmale advantage+(βp-βf)Xffemale disadvantage
Equation 4 gives a more comprehensive decomposition as it provides a breakdown of the unexplained part in terms of male advantage and female disadvantage for each indicator. Furthermore, the value of the explained portion is not subject to bias as it is computed using the coefficients of the pooled estimated, denoted by βP. Additionally, it must be noted that alternative ways of writing the aforementioned expression of differential for endowments and advantages lead to the same decomposition.
Another extension of the O-B decomposition is the decomposition for differences in distribution with quantile regression. The method was proposed by Melly in order to achieve an estimation for an unconditional distribution (Melly, 2005).[3] Following the estimation, the decomposition is done as follows:
q(βmxm)-q(βfxf)total decomposition=q(βmxm)-q(βMmrfxm)residuals+q(βMmrfxm)-q(βfxm)median coefficients+q(βfxm)-q(βfxf)covariates
In the decomposition above for a specific quantile (q), the first portion represents the effect of the differences in residuals, the second portion represents the effect of the differences in the median coefficients of the quantile regression estimate and the third portion represents the effect of the covariates.
In the first two components, βMmrf represents the vector for the median coefficients for the male population with the residual distribution of the female population. The use of median coefficients instead of conditional mean as a central tendency increases the robustness to assumptions regarding distributions. The use of mean makes the regression more sensitive to outliers and extreme values whereas the use of the median makes it less sensitive due to its ordinal ranking nature.
The third portion uses q(βfxm) which is the male characteristics with female coefficients to calculate the explained differential. This is similar to the O-B approach. Furthermore, this component can be interpreted as the rate of return to the individual characteristics on wages.
LITERATURE REVIEW
An extensive amount of literature attempts to explain the gender wage differential. Several studies explore various factors in attempt to explain the gender wage differential on an individual or a macro level.
i. Independent Variables
LEVEL OF INCOME
In the recent decade, several studies have focused on the level of income in order to provide a more detailed explanation for the unexplained gender wage gap. Albrecht et al. (2003) analysed the relation between the gender wage differential and the level of income in Sweden. The study concludes that there exists a ‘glass ceiling’ and ‘sticky floor’ effect in the wages of Sweden. The ‘glass ceiling’ effect indicating a significantly higher wage differential at high income levels and the ‘sticky floor’ indicating significantly higher differentials at very low levels of income.